CA2397090C

A method for encoding images using the classes of neighbouring blocks to determine the prediction method, and an image encoder

Abstract

The invention relates to a method for encoding a digital image,in which method the digital image is divided into blocks (C, L, U, UL, UR). In the method a spatial prediction for a block (C) is performed to reduce theamount of information to be transmitted, wherein at least one prediction method(P1-P13) is defined. In the method a classification is determined for at least oneneighbouring block (L, U) of said block (C) to be predicted according to thecontents of said neighbouring block (L, U), and a prediction method (P1-P13) isselected for the current block (C) on the basis of at least one said classification.

CA2397090C, drawing sheet 1 — Sheet 1 of 13

Term

Term ended

Expired 22 January 2021, 5.7 years ago.

59 claims: 6 independent, 53 dependent

1
CA 02397090 2005-02-11 Claims:1. A method for encoding a digital image in a block-based manner, in which a spatial prediction for a block is performed to reduce an amount of information to be transmitted, the method comprising: 5 - examining pixel values of a neighbouring block of a block to be predicted to determine a classification for the neighbouring block according to the image contents of said neighbouring block;selecting a prediction method for the block to be predicted on the basis of said classification;and 10 - forming a spatial prediction for the block to be predicted using the selected prediction method.
12
A device for encoding a digital image in a block-based manner, the device being arranged to perform a spatial prediction for a block to reduce an amount of information to be transmitted, the device comprising:a classifier arranged to examine pixel values of a neighbouring block of a block to be predicted to determine a classification for the neighbouring block according to the image contents of said neighbouring block;a prediction method selector for selecting a prediction method for the block to be predicted on the basis of said classification;and a prediction estimator for forming a spatial prediction for the block to be predicted using the selected prediction method.
23
An encoder for encoding a digital image in a block-based manner, the encoder being arranged to perform a spatial prediction for a block to reduce an amount of information to be transmitted, the encoder comprising:a classifier arranged to examine pixel values of a neighbouring block of said block to be predicted to determine a classification for the neighbouring block according to the image contents of said neighbouring block;a prediction method selector for selecting a prediction method for the block to be predicted on the basis of said classification;and a prediction estimator for forming a spatial prediction for the block to be predicted using the selected prediction method.
24
A decoder for decoding a digital image in a block-based manner, the decoder being arranged to perform a spatial prediction for a block to reduce an amount of information to be transmitted, the decoder comprising:a classifier arranged to examine pixel values of a neighbouring block of said block to be predicted to determine a classification for the neighbouring block according to the image contents of said neighbouring block ;a prediction method selector for selecting a prediction method for the block to be predicted on the basis of said classification;and a prediction estimator for forming a spatial prediction for the block to be predicted using the selected prediction method.
25
A codec for encoding and decoding digital images in a block-based manner, the codec being arranged to perform a spatial prediction for a block to reduce an amount of information to be transmitted, the codec comprising:a classifier arranged to examine pixel values of a neighbouring block of said block to be predicted to determine a classification for the neighbouring block according to the image contents of said neighbouring block ;a prediction method selector for selecting a prediction method for the block to be predicted on the basis of said classification;and a prediction estimator for forming a spatial prediction for the block to be predicted using the selected prediction method.
37
A method for decoding a digital image in a block-based manner, in which a spatial prediction for a block is performed to reduce an amount of information to be transmitted, the method comprising:examining pixel values of a neighbouring block of a block to be predicted to determine a classification for the neighbouring block according to the image contents of said neighbouring block;selecting a prediction method to be used in forming a spatial prediction for the block to be predicted on the basis of said classification;and forming a spatial prediction for the block to be predicted using the selected prediction method.

Independent claims6

403 paragraphs in 77 sections, as filed

CA 02397090 2005-02-11

A method for encoding images using the classes of neighbouring blocks to determine the prediction method, and an Image encoder

The present invention relates to a method and device for encoding images in a blockbased manner, in which a spatial prediction for a block is performed to reduce an amount of information to be transmitted.

The image can be any digital image, a video image, a TV image, an image generated by a video recorder, a computer animation, a still image, eto. in general, a digital image consists of pixels, are arranged in horizontal and vertical lines, the number of which in a single image is typically tens of thousands. In addition, the information generated for each pixel contains, for instance, luminance information relating to the pixel, typically with a resolution of eight bits, and in colour applications also chrominance information, e.g. a chrominance signal. This chrominance signal generally consists of two components, Cb and Cr, which are both typically transmitted with a resolution of eight bits. On the basis of these luminance and chrominance values, it is possible to form information corresponding to the original pixel on the display device of a receiving video terminal. In this example, the quantity of data to be transmitted for each pixel is 24 bits uncompressed. Thus, the toial amount of information for one image amounts to several megabits. In the transmission of a moving image, several images are transmitted per second. For instance in a TV image, 25 images are transmitted per second. Without compression, the quantity of information to be transmitted would amount to tens of megabits per second. However, for example in the Internet data network, the data transmission rate can be in the order of 64 kbits per second, which makes uncompressed real time image transmission via this network practically impossible.

To reduce the amount of information to be transmitted, a number of different compression methods have been developed, such as the

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

JPEG, MPEG and H.263 standards. In the transmission of video, image compression can be performed either as inter-frame compression, intra-frame compression, or a combination of these. In inter-frame compression, the aim is to eliminate redundant information in successive image frames. Typically, images contain a large amount of non-varying information, for example a motionless background, or slowly changing information, for example when the subject moves slowly. In inter-frame compression, it is also possible to utilize motion compensated prediction, wherein the aim is to detect elements in the image which are moving, wherein motion vector and prediction error information are transmitted instead of transmitting the pixel values.

To enable the use of image compression techniques in real time, the transmitting and receiving video terminal should have a sufficiently high processing speed that it is possible to perform compression and decompression in real time.

In several image compression techniques, an image signal in digital format is subjected to a discrete cosine transform (DCT) before the image signal is transmitted to a transmission path or stored in a storage means. Using a DCT, it is possible to calculate the frequency spectrum of a periodic signal, i.e. to perform a transformation from the time domain to the frequency domain. In this context, the word discrete indicates that separate pixels instead of continuous functions are processed in the transformation. In a digital image signal, neighbouring pixels typically have a substantial spatial correlation. One feature of the DCT is that the coefficients established as a result of the DCT are practically uncorrelated; hence, the DCT conducts the transformation of the image signal from the time domain to the (spatial) frequency domain in an efficient manner, reducing the redundancy of the image data. As such, use of transform coding is an effective way of reducing redundancy in both inter-frame and intra-frame coding.

Current block-based coding methods used in still image coding and video coding for independently coded key frames (intra-frames) use a block-based approach. In general, an image is divided into NxM blocks

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 that are coded independently using some kind of transform coding. Pure block-based coding only reduces the inter-pixel correlation within a particular block, without considering the inter-block correlation of pixels. Therefore, pure block-based coding produces rather high bit rates even when using transform-based coding, such as DCT coding, which has very efficient energy packing properties for highly correlated data. Therefore, current digital image coding standards exploit certain methods that also reduce the correlation of pixel values between blocks.

Current digital image coding methods perform prediction in the transform domain, i.e. they try to predict the DCT coefficients of a block currently being coded using the previous coded blocks and are thus coupled with the compression method. Typically a DCT coefficient that corresponds to the average pixel value within an image block is predicted using the same DCT coefficient from the previous coded block. The difference between the actual and predicted coefficient is sent to decoder. However, this scheme can predict only the average pixel value, and it is not very efficient.

Prediction of DCT coefficients can also be performed using spatially neighbouring blocks. For example, a DCT coefficient that corresponds to the average pixel value within a block is predicted using the DCT coefficient(s) from a block to the left or above the current block being coded. DCT coefficients that correspond to horizontal frequencies (i.e. vertical edges) can be predicted from the block above the current block and coefficients that correspond to vertical frequencies (i.e. horizontal edges) can be predicted from the block situated to the left. Similar to the previous method, differences between the actual and predicted coefficients are coded and sent to the decoder. This approach allows prediction of horizontal and vertical edges that run through several blocks.

In MPEG-2 compression, the DCT is performed in blocks using a block size of 8 x 8 pixels. The luminance level is transformed using full spatial resolution, while both chrominance signals are subsampled. For

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 example, a field of 16 x 16 pixels is subsampled into a field of 8 x 8 pixels. The differences in the block sizes are primarily due to the fact that the eye does not discern changes in chrominance equally well as changes in luminance, wherein a field of 2 x 2 pixels is encoded with the same chrominance value.

The MPEG-2 standard defines three frame types: an l-frame (Intra), a P-frame (Predicted), and a B-frame (Bi-directional). An l-frame is generated solely on the basis of information contained in the image itself, wherein at the receiving end, an l-frame can be used to form the entire image. A P-frame is typically formed on the basis of the closest preceding l-frame or P-frame, wherein at the receiving stage the preceding l-frame or P-frame is correspondingly used together with the received P-frame. In the composition of P-frames, for instance motion compensation is used to compress the quantity of information. Bframes are formed on the basis of a preceding l-frame and a following P- or l-frame. Correspondingly, at the receiving stage it is not possible to compose the B-frame until the preceding and following frames have been received. Furthermore, at the transmission stage the order of the P- and B-frames is changed, wherein the P-frame following the B-frame is received first. This tends to accelerate reconstruction of the image in the receiver.

Intra-frame coding schemes used in prior art solutions are inefficient, wherein transmission of intra-coded frames is bandwidth-excessive. This limits the usage of independently coded key frames in low bit rate digital image coding applications.

The present invention addresses the problem of how to further reduce redundant information in image data and to produce more efficient coding of image data, by introducing a spatial prediction scheme involving the prediction of pixel values, that offers a possibility for prediction from several directions. This allows efficient prediction of edges with different orientations, resulting in considerable savings in bit rate. The method according to the invention also uses contextCA 02397090 2005-02-11 dependent selection of suitable prediction methods, which provides further savings in bit rate.

The invention introduces a method for performing spatial prediction of pixel values within an image. The technical description of this document introduces a method and system for spatial prediction that can be used for block-based still image coding and for intra-frame coding in block-based video coders. Key elements of the invention are the use of multiple prediction methods and the context-dependent selection and signalling of the selected prediction method. The use of multiple prediction methods and the context-dependent selection and signalling of the prediction methods allow substantial savings in bit rate to be achieved compared with prior art solutions.

It is an object of an aspect of the present invention to improve encoding and decoding of digital images such that higher encoding efficiency can be achieved and the bit rate of the encoded digital image can be further reduced.

According to the present invention, this object is achieved by an encoder for performing spatially predicted encoding of image data.

According to a first aspect of the invention there is provided a method for encoding a digital image in a block-based manner, in which a spatial prediction for a block is performed to reduce an amount of information to be transmitted, the method comprising:

examining pixel values of a neighbouring block of a block to be predicted to determine a classification for the neighbouring block according to the image contents of said neighbouring block;

selecting a prediction method for the block to be predicted on the basis of said classification; and forming a spatial prediction for the block to be predicted using the selected prediction method.

According to a second aspect of the invention there is provided a device for encoding a digital image in a block-based manner, the device being arranged to perform a spatial prediction for a block to reduce an amount of information to be transmitted, the device

CA 02397090 2005-10-13 comprising:

a classifier arranged to examine pixel values of a neighbouring block of a block to be predicted to determine a classification for the neighbouring block according to the image contents of said neighbouring block;

a prediction method selector for selecting a prediction method for the block to be predicted on the basis of said classification; and a prediction estimator for forming a spatial prediction for the block to be predicted using the selected prediction method.

According to a third aspect of the invention there is provided an encoder for encoding a digital image in a block-based manner, the encoder being arranged to perform a spatial prediction for a block to reduce an amount of information to be transmitted, the encoder comprising:

a classifier arranged to examine pixel values of a neighbouring block of said block to be predicted to determine a classification for the neighbouring block according to the image contents of said neighbouring block;

According to a fourth aspect of the invention there is provided a decoder for decoding a digital image in a block-based manner, the decoder being arranged to perform a spatial prediction for a block to reduce an amount of information to be transmitted, the decoder comprising:

CA 02397090 2005-10-13

According to a fifth aspect of the invention there is provided a codec for encoding and decoding digital images in a block-based manner, the codec being arranged to perform a spatial prediction for a block to reduce an amount of information to be transmitted, the codec comprising:

According to a sixth aspect of the invention there is provided a method for decoding a digital image in a block-based manner, in which a spatial prediction for a block is performed to reduce an amount of information to be transmitted, the method comprising:

examining pixel values of a neighbouring block of a block to be predicted to determine a classification for the neighbouring block according to the image contents of said neighbouring block;

selecting a prediction method to be used in forming a spatial prediction for the block to be predicted on the basis of said classification; and forming a spatial prediction for the block to be predicted using the selected prediction method.

The invention is based on the idea that to perform spatial prediction of pixel values for a block to be coded, adjacent decoded blocks are examined to determine if there exists some directionality in the contents of the adjacent blocks. This directionality information is then used to classify the blocks. Based on the combination of the classes of the adjacent blocks, the contents (pixel values) of the current block are then predicted using a suitable prediction method. The prediction

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 method is signalled to the decoder. Prediction error information is also sent if it is efficient to do that in a distortion vs. bit-rate sense.

Considerable advantages are achieved with the present invention when compared with solutions of prior art. Using a method according to the invention, it is possible to reduce the amount of information needed when transmitting images in digital format.

In general, the method according to the invention can be applied to block-based still image coding as well as to intra-frame coding in a block-based digital image coder.

In the following, the invention will be described in more detail with reference to the appended figures, in which

Fig. 1 shows the structure of a digital image transmission system,

Fig. 2 illustrates the spatial prediction method of the present invention in the form of a block diagram,

Figs. 3a—3c show an illustration of blocks that are used for prediction according to an advantageous embodiment of the present invention,

Fig. 4 shows the mapping of directionality classes to context classes according to an advantageous embodiment of the present invention,

Figs. 5a—5p show an illustration of pixels that are used for prediction according to an advantageous embodiment of the present invention,

Fig. 6 shows an advantageous bit-stream syntax used in the transmission of displacement information, and

CA 02397090 2002-07-09

WO 01/54416 PCT/FI01/00050

Fig. 7 is a schematic representation of a portable communications device implementing a method according to the invention.

The intra-frame prediction method described in this invention operates in a block-based manner and can be applied to image frames that comprise NxM blocks scanned e.g. row by row from left to right and from top to bottom. It is obvious that other scanning directions can also be used in connection with the present invention. Spatial prediction is performed for each intra-coded block using previously reconstructed blocks in the same frame. The residual error can be compressed using any suitable method, e.g. using DCT, as in current standards. It should also be appreciated that the method according to the invention may be applied equally well to both monochrome and colour images.

The system according to the invention consists of two main parts, as illustrated in Figure 2. Firstly, context-dependent selection 17 of a suitable subset of prediction methods is performed by classifying neighbouring reconstructed blocks. Secondly, a prediction block is constructed 18 using one of the prediction methods in the selected subset and the prediction method is signalled to decoder.

Context-dependent selection of a prediction method subset comprises directionality classification of possible neighbouring blocks, mapping of directionality classes to context classes and context-dependent selection of an appropriate prediction method subset.

In the following, the transmission and reception of digital image frames in a transmission system is described with reference to the digital image transfer arrangement presented in Figure 1. The current frame arrives at the transmission system 1 as input data 2 provided, for example, as the output of a digital video camera. The current frame may be provided in its entirety (i.e. a complete frame comprising NxM image blocks), in which case the frame is stored, or the transmission system 1 may receive the input data block by block. The blocks of the frame are directed one by one to a summer 4, where prediction error of a block is calculated e.g. by subtracting a block of the frame from a

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 predicted block. The prediction error is coded in a coder 5 and decoded in a decoder 6. In summer 7 the decoded prediction error is summed with predicted blocks and the result is saved in a frame memory 8. The prediction estimator 3, where spatial prediction is performed according to the method of the invention, receives blocks to be used with prediction from the frame memory 8.

In order to form a new prediction block, the prediction estimator 3 examines, if there exists some directionality in possible neighbouring blocks of the current block. This scheme is illustrated in Figure 3a. The reference C denotes the current block, the reference L denotes a first neighbouring block of the current block and the reference U denotes a second neighbouring block of the current block. In this advantageous embodiment of the invention, the first neighbouring block is to the left of the current block C and the second neighbouring block is above the current block C. If the scanning order is different from left to right and from top to bottom, the first neighbouring block L and the second neighbouring block U are not necessarily to the left of and above the current block C, respectively. The neighbouring blocks L, U are blocks adjacent to the current block C which have already been reconstructed. In some embodiments of the invention more than two blocks can be classified and used to select the prediction method for the current block

C. However, in the following description of a preferred embodiment of the invention, a maximum of two neighbouring blocks L, U are classified for each block C under examination. Furthermore, the classification is performed only if a neighbouring block L or U exists. If a current block does not have any neighbouring blocks, it is treated as Non-intra during context-dependent selection of prediction methods, as will be explained further later in the text.

Prediction can also be implemented in such a way that it is performed using only already reconstructed intra-coded blocks. In this case, all blocks other than intra-coded blocks are treated as ”Non-lntra”.

The first neighbouring block L and the second neighbouring block U are classified according to the directionality of image details inside the

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 block. As illustrated in Figure 2, directionality classifier 19 analyses the directionality of the neighbouring blocks using pixel value gradients. As a result, each neighbouring block is mapped 20 into an output class. In an advantageous embodiment of the invention there are 11 such output classes, but it is obvious that the number of output classes may vary. Advantageously, the output classes consist of 8 directionality classes DO - D7 corresponding to edge orientations k«22.5°, k = 0, 1, ... , 7 and 3 non-directional classes D8 - D10 corresponding to flat, smooth texture and coarse texture blocks. In alternative embodiments of the invention, the number of directionality classes and the way in which they are defined may vary.

In the system of figure 1, the prediction estimator 3 first examines if the first neighbouring block L and/or the second neighbouring block U exist. If either one of these blocks does not exist, that neighbouring block is defined as a CO block (”Non-lntra”), i.e. the current block C is on the edge or in a corner of the frame, or on the edge or in a corner of an area consisting of Intra blocks. Then, the prediction estimator 3 selects a suitable prediction method for the current block C, as described later in this description. Otherwise, the prediction estimator 3 calculates gradient information relating to the block or blocks L, U.

There are many suitable methods for calculating the gradient information. In the following, one advantageous method is described. First, average absolute directional gradients gk, k = 0, 1, ... , 7 of a block L, U are defined as

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

N—X N-2 go = max /

1, Σ ΣΙΖ(*’ yï-Kx+l-, y)| y=0 x=Q

N—2N—X gx gx g5 = (N-iy 1 (N-iy _1__

N(N-Ï) 1

-max

-max x’ Σ Σ lz(*’ y) “ 2 (z(* -1, y)+/(x -1, y +1))| y=0 x=X ( N-2N-X A Χ’ΣΣΙ7^’ y) — 7(x—1, y+ 1)| y=0 j=l

-max max

1, ΣΣΙΖ^> y)“ y + l) + /(x, y + l))| >'=0 x=X j ( N—2N—X V

1,ΣΣΙΖ(Λ> yWO,y+i)|

7=0 x=0 g6 = gl = {N-iy (Λ/-1)2 (w-iy

-max ( N-2 N-2 max

-max Χ>ΣΣΙ IV, y) - {(/(*, y+1) + J(x+1, y+1))|

7=0 :c=0

N-2N-2 ι,ΣΣΙα*. y)-/(x+l, y + l)|

7=0 x-0

y)-j(/(x+l, y) + /(x+l, y+1))|

7=0 -τ=0 (1) where N is the size of the block and l(x,y) represent the pixel intensity values. Indices x and y refer to the co-ordinates of a pixel inside the block and k represents edge orientations. The prediction estimator 3 calculates the gradient values gk according to the formulae above.

Using the gradient values gk, gradient ratios rk, k = 0, 1, ... , 7 are defined as the ratio between the gradient value in a certain direction and gradient value in the orthogonal direction:

(2)

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

Based on the absolute gradient values gk and gradient ratios rk defined in (1) and (2), classification of the block is performed, advantageously according to the following classification steps 1—12 using some numerical values as thresholds. This classification process classifies each of the neighbouring blocks into one of a first set of block types DO—D10. The present invention is not limited to the values used in the algorithm, but the values used in the algorithm in the following steps are preferred. The method can also be applied to any block size.

In this advantageous embodiment of the invention the classification phase comprises 13 steps, but it is obvious that the classification may comprise also different number of steps.

Step 1

In this step the flatness of the block is checked. Prediction estimator 3 calculates gradient values g0 and g4. These correspond to gradient values for horizontal (0°) and vertical (90°) image details. If both g0 < 2.0 and g4 < 2.0, the block is classified as class D8 and the initial classification process terminates. Otherwise, classification step 2 is performed.

Step 2

In this step a further check for flatness of the block is performed. The rest of the gradient values gk are calculated, and the maximum gradient value gmax = max{gk} is determined. The maximum gradient value gmax is compared with 2.5. If gmax < 2.5 the block is classified as class D8 and the initial classification process terminates. Otherwise, the method continues from step 3.

Step 3

In step 3 a check for clear directionality is performed. The gradient ratios rk are calculated and the minimum gradient ratio rmin = min{rk} is determined. When the minimum gradient ratio is found, the corresponding index kmjn is defined. If rmin 0.15 the block is classified to corresponding class Dkmin and the method continues from step 12, otherwise the method continues from step 4.

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

Step 4

In step 4 a check for texture is performed. The minimum gradient ratio rmin is compared with 0.6. If rmin > 0.6 the method continues from step 13, otherwise the method continues from the next step.

Step 5

In step 5 the two smallest gradient ratios are checked to determine if they are clearly distinct. The gradient ratios rk are sorted in increasing order r(0) < r(1) < r(2) ...<r(7). Also the gradient ratio indices are reordered according to the sorted order k(0), k(1), k(2), ... k(7). If rd)-fio) <l(r(2)-rd)) fbe sixth classification step is performed next, otherwise the method continues from the 10th classification step.

Step 6

In step 6 the smallest gradient ratio is checked to determine if it corresponds to directionality class D2 or D6 and the smallest gradient ratio is small enough. The prediction estimator 3 first examines, whether the index of the gradient ratio r(0) is either 2 or 6, wherein the first gradient ratio r(0) is compared with 0.6. If r(0) e {rjfc = 2,6} and r(0) <

0.6, the block is classified as corresponding to class Dk(0) and the method continues from step 12. Otherwise the method continues from step 7.

Step 7

In step 7 the prediction estimator 3 first examines if the index of the second gradient ratio r(1) is either 2 or 6, wherein the first gradient ratio r(0) is compared with 0.6. If r(1) e {rj k = 2,6} and r(0) < 0.6 the block is classified as corresponding to class Dk(1) and the method continues from the step 12, otherwise the method continues from Step 8.

Step 8

In step 8 the smallest gradient ratio is checked to determine if it corresponds to directionality class D1, D3, D5 or D7 and the smallest gradient ratio is small enough. The first gradient ratio η0) is compared

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 with 0.5. If r{0) E{rk | k = 1,3,5,7} and r(0) < 0.5 the block is classified as corresponding to class Dki0) and the method continues from step 12, otherwise the method continues from step 9.

Step 9

In step 9 the second gradient ratio is checked to determine if it corresponds to directionality class D1, D3, D5 or D7 and the smallest gradient ratio is small enough. The first gradient ratio r(0) is compared with 0.5, if ?-(1) e{rk | k = 1,3,5,7}. If r(0) < 0.5 the block is classified as corresponding to class Dk(1) and the method continues from step 12. Otherwise the method continues from step 10.

Step 10

Directionality is not yet found, therefore a (somewhat) higher threshold value compared with the threshold value used in Step 3 can be used to check the directionality. This means that a more uncertain examination is performed. Step 10 uses the values of threshold ΤΊ defined in Table 1, below. The values for T-ι are compared with the first gradient ratio. If r(0) < T-ι as defined in Table 1, the block is classified as corresponding to class Dk(o) and the method continues from step 12. Otherwise the method continues from step 11.

<td> Orientation Relation for r«n</td><td> fi</td>

Table 1

Step 11

Directionality is not yet found, therefore in step 11 the three smallest gradient ratios are checked to determine if they are neighbours and if the smallest gradient ratio is in the middle. In that case a still higher threshold value compared with the threshold value used in Step 3 can be used to check the directionality. This means that a more uncertain examination is performed. Step 11 uses the values of threshold T2

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 defined in Table 2, below. Then, if the directionalities corresponding to the second r(1) and the third gradient ratios r(2) are the closest neighbours for the directionality corresponding to the first gradient ratio r(0) and r(0) < T2 as defined in Table 2, the block is classified as corresponding to class Dk(0) and the method continues from step 12. Otherwise the method continues from step 13.

<td> Orientation Relation for rf0)</td><td> T2</td>

Table 2

Step 12

Step 12 performs a check that classification is really based on an edge in the image with a certain orientation rather than texture. Step 12 uses the values of threshold T3 defined in Table 3, below. In Table 3 values for only two possible block sizes (8x8, 4x4) are shown, but in practical embodiments other block sizes can also exist, wherein respective values for T3 are defined. In step 12 the minimum gradient value gmin = min{gk} is examined. Depending on the classification and the size of the block, the threshold T3 is chosen from Table 3. If gmin < T3 the initial classification process terminates. Otherwise the method continues from step 13.

<td> Classification of the Block</td><td> T3 for 4x4 Block</td><td> T3for 8x8 Block</td>

Table 3

Step 13

Step 13 performs a check whether texture is smooth or coarse. The maximum gradient value gmax is compared with 10.0. If gmax^ 10 0 the

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 block is classified as D9. Otherwise, the block is classified as D10. Step 13 is not necessarily needed, if both smooth and coarse texture are mapped into the same context class.

Next the selection 21 of a suitable prediction method is performed for the current block C. In a preferred embodiment of the invention, the selection phase is preceded by a mapping phase. The purpose of the mapping is to reduce the memory consumption of the implementation. Some of the directionality classes can be mapped together. The classes resulting from the mapping phase are called context classes and they are referred to with references C1 - C6. In the preferred embodiment of the invention, the diagonal classes are combined to two alternative classes, one for bottom-left to top-right diagonality and the other for top-left to bottom-right diagonality.

Mild and steep diagonal classes D5, D6 and D7 are mapped to the first diagonal context class C4. Similarly, classes D1, D2 and D3 are mapped to the second diagonal context class C2. Further, the smooth texture class D9 and coarse texture class D10 are mapped together to produce texture context class C6. This mapping is illustrated in Figure

In addition to the 6 context classes C1—C6 there is one further context class CO used for Non-lntra” blocks. In general, a ”Non-lntra” block is a block that does not exist, i.e. when block C is at an image boundary. If the prediction is implemented in such a way that only intra-coded blocks are used as a reference, the definition of a ”Non-lntra” block is extended to those blocks that are not intra-coded.

In the preferred embodiment of the invention there are a total of 13 different prediction methods, which are depicted in Figures 5a—5p for 8x8 blocks. Prediction methods for other block sizes and context classes can be derived in a similar fashion. In each case, prediction is performed in a causal manner, using neighbouring reconstructed intracoded blocks L, U, UL, UR as a reference. The region used for prediction depends on the prediction method, as depicted in Figures 3a

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 and 3b, where block C is the current block to be coded. In the case of prediction methods P1—P12, the region from which blocks may be used for prediction is the area covered by four neighbouring blocks L, UL, U and R as shown in Figure 3b. For prediction method P13, this region is larger, as depicted in Figure 3c. It should be appreciated that in other embodiments of the invention, the number of prediction methods, the blocks used as prediction references, as well as the pixels within those blocks used to perform prediction, may vary.

In an advantageous embodiment of the method according to the invention, a subset of prediction methods for each context class combination is defined and the prediction methods are prioritized (ranked) in each subset. Then, the prediction method used to predict the content of the current block C is selected from a subset of prediction methods. The prediction methods within a subset differ from each other and correspond to those prediction methods that are most likely to provide an accurate prediction for block C, in the event of particular classifications being obtained for neighbouring blocks like L and U. One advantageous definition for the subsets is presented in Table 4 below.

Effectively, the results of context classification for the first neighbouring block L and second neighbouring block U are combined, i.e. both taken into consideration when selecting a prediction method for block C. The subset of prediction methods is selected from Table 4 according to the context information of the neighbouring blocks L, U. Each row of Table 4 defines the prediction method subset for a certain pair of context classes for neighbouring blocks L, U and the priority (rank) of the prediction methods in the subset. Ranking is used to simplify the context-dependent signalling of the prediction methods, as described later in this description. For example, if the first neighbouring block L is classified into context class C2 and the second neighbouring block U is classified into context class C4, the subset for this combination comprises prediction methods Pl, P9, P5, P13, P7 and P6 (in ranking order). The prediction estimator 3 further selects the most appropriate prediction method from this subset, as detailed later in this description.

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

<td colspan="2"></td><td colspan="6"> Rank of Prediction Methods</td>

<td> L Class</td><td> U Class</td><td> Rank 1</td><td> Rank 2</td><td> Rank 3</td><td> Rank 4</td><td> Rank 5</td><td> Rank 6</td>

Table 4

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

In the following, the defined prediction methods are described in more detail.

Prediction method P1

Prediction method P1 predicts the average pixel value of block C from the average pixel values of blocks L, UL, U and UR. The average pixel values dL, dUL and dU of the reconstructed blocks L, UL, and U are calculated as the integer division defined as ( N-1,N-1

HN‘ (3) where N is the size of the block, l(x,y) represents the pixel intensity values and 7/ denotes division with truncation to integer value. The average pixel value dC of block C is predicted according to following set of rules (which are written below in the form of pseudo-code):

if all blocks L, U and UL exist, then if dL = dU = dUL then dC = dUL else if dUL = dU then dC = dL else if dUL = dL then dC = dU else if dL = dU then if chrominance prediction then else if | dUL - dL | < 4 then else dC = dL else if dUL < dL < dU then dC = dU else if dUL < dU < dL then dC = dL else if dU < dL < dUL then dC = dU else if dL < dU < dUL then dC = dL dC = dL dC = s(dL + dU - dUL) else if dL < dUL < dU OR dU < dUL < dL then dC = s(dL + dU - dUL) else if blocks L and U exist then dC = (dL + dU +1) // 2 else if blocks L and UL exist then dO = dL else if blocks U and UL exist then dC = dU

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 else if block L exists then else if block U exists then else if block UL exists then else dC = dL dC = dU dC = dUL dC = p where p is a value that is in the middle of the possible pixel value range, e.g. 128, 7/” denotes division with truncation and s is a clipping function that restricts the values to the possible range of pixel values, e.g. between 0 and 255 in a system that uses an 8-bit representation of luminance/chrominance values. As a result, the prediction block for C is filled with pixels having a constant value given by dC. Prediction method P1 is illustrated in Figure 5a.

Prediction method P2—P4

Prediction methods P2 through P4 predict diagonal shapes in block C by extending image details from the upper right direction into block C. Prediction is performed by copying reference pixel values at the boundaries of blocks U and UR into block C, as depicted in Figures 5b, 5c, 5d, respectively. Reference pixels that are marked in grey are connected to one or more predicted pixels. The connection is marked as line with dots to indicate connected predicted pixels, the value of the reference pixel is copied to all connected predicted pixels.

Since one or more reference blocks might be unavailable, i.e. their context class may be CO, prediction is performed according to following rules.

Rule 1

If both blocks, U and UR, are classified into one of classes C1 - C6, pixel prediction is performed as shown in Figures 5b, 5c and 5d respectively. For prediction method P2 (Figure 5b), pixels without any corresponding reference pixel in block UR are advantageously allocated the value of the rightmost reference pixel in block UR.

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

Rule 2

If block U is classified into one of classes C1 - C6 and block UR is classified as CO, pixel prediction is performed as shown in Figures 5b, 5c and 5d for pixels that have a reference pixel in block U. The rest of the pixels are advantageously set to the value of the pixel in the lower right corner of the reference block U.

Rule 3

If block U is classified as CO, the current block C is advantageously filled with pixels having a constant value that is substantially in the middle of the possible dynamic range of pixel values, e.g. 128 (in a system, that uses an 8-bit representation of luminance/chrominance values).

Prediction method P5 and P9

Prediction methods P5 (Figure 5e) and P9 (Figure 5i) predict vertical and horizontal shapes in the current block C by extending image details into the current block C, either from above or from the left. Depending on the selected method (P5 or P9), the reference pixel values at the boundary of either block U or L are copied to the current block C as depicted in Figures 5e and 5i.

If the context class of the reference block is CO then the current block C is advantageously filled with pixels having a constant value that is substantially in the middle of the possible dynamic range of pixel values, e.g. 128 (in a system, that uses an 8-bit representation of luminance/chrominance values).

Prediction method P6, P7 and P8

Prediction methods P6, P7 and P8 predict diagonal shapes in the current block C by extending image details from the upper left direction into the current block C as depicted in Figures 5f, 5g and 5h, respectively. Prediction is performed by copying reference pixel values at the boundaries of blocks L, UL and U into the current block C according to following rules.

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

Rule 1

If all blocks L, UL and U are classified into one of classes C1 - C6, the pixel prediction for the current block C is performed as illustrated in Figures 5f, 5g and 5h.

Rule 2

If blocks UL and U are classified into one of classes C1 - C6 and block L is classified as CO, pixel prediction for the current block C is performed as shown in Figures 5f, 5g and 5h for those pixels of the current block C that have a reference pixel in blocks UL and L. The remaining pixels in the current block C are advantageously assigned the value of the pixel in the lower left comer of the reference pixel area in block UL.

Rule 3

If blocks L and UL are classified into one of classes C1 - C6 and block U is classified as CO, pixel prediction for the current block C is performed as shown in Figures 5f, 5g and 5h for those pixels of the current block C that have a reference pixel in blocks L and UL. The remaining pixels in the current block C are advantageously assigned the value of the pixel in the upper right comer of the reference pixel area in block UL.

Rule 4

If blocks L and U are classified into one of classes C1 - C6 and block UL is classified as CO, pixel prediction for the current block C is performed as shown in Figures 5f, 5g and 5h for those pixels of the current block C that have a reference pixel in blocks L and U. Pixels with reference pixel in block UL are predicted as shown in Figures 5n, 5o and 5p. In case of method P7, the predicted pixel value is the average of the two reference pixel values rounded to the nearest integer value, as indicated in Figure 5o.

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

Rule 5

If block L is classified into one of classes C1 - C6 and blocks UL and U are classified as CO, pixel prediction for the current block C is performed as shown in Figures 5f, 5g and 5h for those pixels of the current block C that have a reference pixel in block L. The remaining pixels in the current block C are advantageously assigned the value of the pixel in the upper right comer of the reference pixel area in block L.

Rule 6

If block UL is classified into one of classes C1 - C6 and blocks L and U are classified as CO, pixel prediction for the current block C is performed as shown in Figures 5f, 5g and 5h for those pixels of the current block C that have a reference pixel in blocks UL. Pixels of the current block C that have a reference pixel in block L are advantageously assigned the value of the lower/left reference pixel in block UL. Pixels of the current block C that have a reference pixel in block U are assigned the value of the upper/right reference pixel in block UL.

Rule 7

If block U is classified into one of classes C1 - C6 and blocks L and UL are classified as CO, pixel prediction for the current block C is performed as shown in Figures 5f, 5g and 5h for those pixels of the current block C that have a reference pixel in block U. The remaining pixels of the current block C are advantageously assigned the value of the pixel in the lower left corner of the reference pixel area in block U.

Rule 8 if all blocks L, UL and L are classified as CO, the current block C is advantageously filled with pixels having a constant value that is substantially in the middle of the possible dynamic range of pixel values, e.g. 128 (in a system, that uses an 8-bit representation of luminance/chrominance values).

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

Prediction method P10, P11 and P12

Prediction methods P10 through P12 predict diagonal shapes in the current block C by extending image details from the left into the current block C as depicted in Figures 5j, 5k and 5I, respectively. Prediction is performed by copying reference pixel values at the boundary of blocks L into the current block C according to following rules.

Rule 1

If block L is classified into one of classes 01 - 06, the pixel prediction for the current block C is performed as illustrated in Figures 5j, 5k and

5I. Pixels of the current block C without reference pixel in block L are advantageously filled with the value of the pixel in the lower right corner of the reference pixel area.

Rule 2

If block L is classified as CO, the current block C is advantageously filled with pixels having a constant value that is substantially in the middle of the possible range of pixel values, e.g. 128 (in a system, that uses an 8-bit representation of luminance/chrominance values).

Prediction method P13

Prediction method P13 predicts the content of the current block C from the neighbouring image content by examining if there exists a range of pixels having values which substantially corresponds to the pixel values of the current block C. The prediction of the current block C is performed by copying reconstructed pixel values from a reference block B that is inside a search range SR as depicted in Figure 5m. Search range SR is defined by lists of horizontal (x) and vertical (y) displacements. Each pair of horizontal displacement and corresponding vertical displacement values (x, y) defines a displacement vector between the coordinates of upper left corner of the current block C and upper left corner of the reference block B. Prediction is allowed only for those displacements corresponding to reference block B that is completely inside the reconstructed part of the frame. Examples of displacement pairs using 512 displacements for 8x8 blocks are presented in Tables 9a and 9b. In this example the scanning order of

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 the tables is from top-left to bottom-right row by row. In alternative embodiments of the invention, the search range may be different from that depicted in Figure 5m and / or the displacement between the reference block B and the current block may be defined differently.

The list of allowed displacements is known to both the encoder and the decoder, allowing context-dependent signalling of the selected reference block location.

There are many alternative ways to select the prediction method from a subset of prediction methods. For example, a cost function can be defined in order to evaluate the effectiveness of the different prediction methods of the subset to be used. The cost function may be calculated on the basis of information concerning the error incurred when predicting a current block C using a particular prediction method. This error denotes differences between actual pixel values and reconstructed pixel values. Typically, the error values for each pixel in the current block C are squared and summed together to produce a squared error measure for the whole block. The cost function may also comprise information concerning the number of bits, i.e. the bit rate needed to transfer the information to the receiver. The elements of the cost function, particularly the bit rate, can also be weighted to emphasize them. One example of a cost function is:

Cx = D + ÀR, (4) where cost Cx is defined as a weighted sum of distortion D and rate R associated with each of the prediction methods and λ is the weighting factor. If the transmission system is band limited, the weight value is typically larger than if the bandwidth is wider. The values for formula (4) are calculated for different prediction methods and preferably that prediction method which yields the smallest value for the cost function is selected.

Additionally, the prediction error information can also be coded prior to transmission to the receiver. Advantageously, there is a subset of

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 coding methods defined for each prediction method. Specifically, the coding method could be chosen to minimise the number of bits required to encode the prediction error. For example, the effectiveness (bit rate) of the coding method is examined.

If the prediction error is relatively small, it may not be necessary to transmit the prediction error information at all.

Referring once more to Figures 1 and 2, once a suitable prediction method has been selected for predicting a current block C, the prediction estimator 3 performs spatial prediction 22 according to the selected prediction method. The prediction estimator 3 directs the reconstructed block to summer 4 where the reconstructed block is subtracted from the actual contents of the current block C to produce prediction error information for the current block.

The encoder 1 sends 23 the information about the selected prediction method to the multiplexer 9, which is accompanied by displacement information if method P13 is used. Advantageously, the selected prediction method is indicated by its rank in the subset of prediction mehtods appropriate for the particular combination of neighbouring blocks (U, L) in question. Encoding of the information is advantageously performed using variable length coding.

The information is further transmitted to the receiver 10, where the demultiplexer 11 demultiplexes the received information. In the receiver 10 the prediction information is directed to the predictor 16. The receiver 10 also comprises a frame memory 14, where the previously reconstructed blocks are saved. When a new coded block arrives at the receiver, the predictor 16 performs the classifying steps for the neighbouring blocks U, L of the received current block C to classify them into directionality classes, as previously described. Then the predictor 16 carries out the mapping of classification information into context classes C1 - C6. After that the predictor 16 also examines the rank of the prediction method. The receiver 10 contains the information of the Table 4 and 5, wherein the predictor 16 can determine the

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 correct prediction method according to the context class combination and the rank.

When the prediction method has been determined, the predictor 16 can reconstruct the current block C and save it to the frame memory 14. In a situation where prediction error information is also received, that information is first decoded in the decoder 12, if necessary, and combined with the pixel values of the reconstructed block C. Now the current block C is ready to be directed to the output 15 of the receiver.

If the prediction method of the current block C is P13, the reconstruction of current block C is performed in a slightly different manner. In this case, the receiver 10 also has to decode the displacement information, wherein the displacement information is used to copy the pixel values of the current block C from previously reconstructed pixel values in the frame memory 14.

Signalling of the prediction method is advantageously based on the context-dependent codes defined in Table 5. After selecting the appropriate prediction method, the encoder 1 sends a variable length codeword that corresponds to the rank of the selected prediction method in the context-dependent subset. Advantageous examples of variable length codewords representing each prediction method rank are listed in Table 5. For example, if the first neighbouring block L is classified into context class C3 and the second neighbouring block U is classified into context class C1, and the prediction method P9 is selected from the subset of the prediction methods for this combination, the respective rank is 2. Then, the codeword which corresponds this rank is ”01”.

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

<td> Rank</td><td> Code</td><td> Length</td>

Table 5

The receiver 10 is aware of the contents of Table 4, i.e. it knows which prediction method corresponds to each of the ranks in every possible context (combination of classes for the neighbouring blocks L and U). Since the receiver 10 can derive the same context information as the prediction estimator 3, receiver 10 can associate the rank represented by the received codeword to correct prediction method and perform the spatial prediction for block C according to the method.

In an advantageous embodiment of the invention the signalling of horizontal and vertical displacements associated with prediction method P13 is performed as follows:

Step 1

Those pairs of horizontal and vertical displacements (X(i), Y(i)) that correspond to reference blocks B lying partially or entirely outside the frame are eliminated from the ordered list given in Tables 9a, 9b. The number of valid pairs is denoted by Nv and the ordered list of valid pairs which are retained after the elimination is denoted by Lv.

Step 2

The rank r (which is one of 1, 2, ..., Nv) corresponding to the chosen block B within the list Lv created in Step 1 is calculated.

Step 3

Based on the value of rank r determined in Step 1 the value index! is calculated according to Table 6.

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

Step 4

The value index2 = r - OffsetLowiindex^ is calculated using the values listed in Table 6.

<td colspan="2"> Range for rank r</td><td> index-,</td><td> OffsetLow (index/</td><td> OffsetHigh (index/</td><td> AuxLength (index/</td>

Table 6

Step 5

Next, a variable bits is calculated as follows. If Nv < OffsetHigh(index-i), the value for the variable bits is computed advantageously using the formula bits = [Iog2(1 + Nv - OffsetLow(indexl))], where [x] denotes the nearest integer > x. Otherwise, bits = AuxLength(indexl).

Step 6

Depending on the value of Nv the variable whose sub-script is index! is encoded using the corresponding Variable Length Coding given in Table 7 and Table 8. This codeword is transmitted to the decoder, which is illustrated with block CW1 in Figure 6.

Step 7

If the variable bits is nonzero the binary representation of index2 is encoded using a number of bits corresponding to the value of variable

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050 bits and this codeword is transmitted to the receiver, which is illustrated with block CW2 in Figure 6.

<td colspan="3"> Nv in range</td><td colspan="3" rowspan="2"> Nv in range 17,...,32</td><td colspan="3" rowspan="2"> Nv in range 33, ..., 64</td>

<td> Symbol</td><td> Length</td><td> Code</td><td> Symbol</td><td> Length</td><td> Code</td><td> Symbol</td><td> Length</td><td> Code</td>

Table 7

<td colspan="3"> Nv in range 65.....128</td><td colspan="3"> Nv in range 129, ... ,256</td><td colspan="3"> Nv in range 257, ... , 512</td>

<td> Symbol</td><td> Length</td><td> Code</td><td> Symbol</td><td> Length</td><td> Code</td><td> Symbol</td><td> Length</td><td> Code</td>

Table 8

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

X[512] =

Table 9a

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

Y[512] =

Table 9b

CA 02397090 2002-07-09

WO 01/54416

PCT/FI01/00050

Since the decoder can derive the ordered list of valid displacement vectors, it can associate the rank represented by the received codeword with the correct displacement vector.

The block carrying out prediction method according to the invention is particularly advantageously implemented in a digital signal processor or a corresponding general purpose device suited to processing digital signals, which can be programmed to apply predetermined processing functions to signals received as input data. The measures according to the invention can be carried out in a separate signal processor or they can be part of the operation of such a signal processor which also contains other arrangements for signal processing.

A storage medium can be used for storing a software program comprising machine executable steps for performing the method according to the invention. In an advantageous embodiment of the invention the software program can be read from the storage medium to a device comprising programmable means, e.g. a processor, for performing the method of the invention.

A mobile terminal 24 intended for use as a portable video telecommunications device and applying the method according to the invention comprises advantageously at least display means 25 for displaying images, audio means 26 for capturing and reproducing audio information, a keyboard 27 for inputting e.g. user commands, a radio part 28 for communicating with mobile network, processing means 29 for controlling the operation of the device, memory means 30 for storing information, and preferably a camera 31 for taking images.

The present invention is not solely restricted to the above presented embodiments, but it can be modified within the scope of the appended claims.

Contents77

38 members in 14 offices

This recordMember, by publication datePriority claim4 more offices are in the list below

Priority claims9

Priority claims
Document	Office	Kind	Date
20000131	Finland	A	2000-01-21
20000131	Finland	A	2000-01-21
20000131	Finland	–	2000-01-21
0100050	Finland	W	2001-01-22
0100050	Finland	W	2001-01-22
20000131	–	–	–
FI20000000131	–	–	–
PCTFI01000050	–	–	–
WO2001FI00050	–	–	–

Members38

Family members, oldest first
Document	Office	Kind
FI20000131A0	Finland	A0	2000-01-21
FI20000131A	Finland	A	2001-07-22
CA2397090A1	Canada	A1	2001-07-26
CA2572566A1	Canada	A1	2001-07-26
CA2614571A1	Canada	A1	2001-07-26
WO0154416A1	World Intellectual Property Organization (WIPO)	A1	2001-07-26
AU3027601A	Australia	A	2001-07-31
US2001017942A1	United States of America	A1	2001-08-30
KR20020077888A	Republic of Korea	A	2002-10-14
EP1249132A1	European Patent Office (EPO)	A1	2002-10-16
BR0107706A	Brazil	A	2002-11-19
ZA200205506B	South Africa	B	2003-03-26
CN1418436A	China	A	2003-05-14
JP2003520531A	Japan	A	2003-07-02
HK1054288A	Hong Kong, China	A	2003-11-21
US6907142B2	United States of America	B2	2005-06-14
CN1214647C	China	C	2005-08-10
CN1658677A	China	A	2005-08-24
US2005254717A1	United States of America	A1	2005-11-17
HK1054288B	Hong Kong, China	B	2006-02-03
FI116819B	Finland	B	2006-02-28
HK1080653A	Hong Kong, China	A	2006-04-28
CA2397090CThis record	Canada	C	2007-03-13
KR100714355B1	Republic of Korea	B1	2007-05-02
US7295713B2	United States of America	B2	2007-11-13
JP2008054335A	Japan	A	2008-03-06
CA2572566C	Canada	C	2008-03-25
US2008247657A1	United States of America	A1	2008-10-09
CN100473167C	China	C	2009-03-25
HK1080653B	Hong Kong, China	B	2009-07-10
US7567719B2	United States of America	B2	2009-07-28
EP1249132B1	European Patent Office (EPO)	B1	2011-04-27
AT507678T	Austria	T	2011-05-15
ATE507678T1	Austria	T1	2011-05-15
DE60144513D1	Germany	D1	2011-06-09
JP2012170122A	Japan	A	2012-09-06
CA2614571C	Canada	C	2014-03-25
BRPI0107706B1	Brazil	B1	2017-04-04

2 legal events, as the office reported them to INPADOC

Over the term

Point at a mark for the event

Events

Legal events, newest first
	Event	Code
2021-03-28	ExpiryMKEX	MKEX
2002-07-09	Examination requestEEER	EEER