Interframe coding system and method

1. Field of the Invention

This invention relates to an interframe coding system according to the pre-characterizing part of claim 1 and to an interframe coding method according to the pre-characterizing part of claim 4.

2. Discussion of the Related Art

It is a common practice to remove redundant components in an image signal for a highly efficient coding. Especially with a dynamic image signal, an interframe coding is one of the preferable and typical arts. The interframe coding is a coding technique which codes a difference between a newly input uncoded image signal and a previously coded image signal.

Fig. 7 is a block diagram of a conventional interframe coding apparatus disclosed in Japanese Unexamined Patent Publication No. 208382/1988. The interframe coding apparatus according to the figure has a frame memory 1, a motion-vector detector 2, a subtractor 3, an encoder 4, a local decoder 5, an adder 6, a filter 7 and a filter controller 8. The frame memory 1 stores an image signal of the previous frame.

Operation of the conventional interframe coding system with a local decoding loop is now described with reference to the Fig. 7.

The input image signal 12 is compared with the image signal of the previous frame 11 stored in the frame memory 1 by a block-matching technique in the motion vector detector 2. The motion vector detector 2 detects the quantity and direction of motion of the input image signal 12 and outputs a motion vector signal 13. The frame memory 1 outputs a motion compensation predictive signal 14 based upon the motion vector signal 13. The subtractor 3 subtracts the motion compensation predictive signal 14 from the input image signal 12 to output a predictive error signal 15 or a difference signal.

The predictive error signal 15 is coded by quantization at the encoder 4, and a coded error signal 16 is output. The coded error signal 16 is decoded in the local decoder 5, and a local decoded error signal 17 is output. The local decoded error signal 17 is added to the motion compensation predictive signal 14 in the adder 6, and a local decoded signal 18 is output. The local decoded signal 18 is filtered by the filter 7 to eliminate higher frequency components in the signal. The filter smooths the local decoded signal and outputs a smoothed local decoded signal 19. Filtering is controlled by a control signal 20 output by the filter controller 8. The filter control signal 20 controls the filter based upon the motion vector signal 13.

The coded error signal 16 and the motion vector signal 13 are transmitted via a transmission line to an external decoding system.

Image coding is generally processed by the unit or block of 16x16 or 8x8 pixels of an image signal.

With reference to other conventional coding techniques, the filter 7 can be placed after the frame memory 1 in the local decoding loop instead of after the adder 6 as shown in Fig. 7. Further, filtering can be accomplished by an intra-block filter which processes pixels within a block, and by an inter-block filter which processes pixels involving the pixels in the neighboring blocks. Furthermore, a motion detection can be achieved with a smaller unit of a pixel than a full pixel. This is designed to detect the optimal block of pixels in the previous frame which match a block of pixels in the input image signal 12. These conventional techniques can contribute to eliminate higher frequency components or redundancy in the input image signal based upon the quantity of motion. Thus, they are effective for removing noise in the signal. Consequently, coding efficiency can be improved greatly with these conventional arts.

An intraframe coding with a local decoding loop is processed in the following manner according to the conventional coding with reference to Fig 7. The input image signal 12 is directly coded by quantization in the encoder 4, where the coded error signal 16 is output. The coded error signal 16 is decoded in the local decoder 5, where the local decoded error signal 17 is output. The local decoded error signal 17 is stored in the frame memory 1. The coded error signal 16 is transmitted via a transmission line to an external decoding system.

The conventional interframe coding generally controls filtering based upon the motion vector, whereby an image signal can be filtered based upon the quantity of motion. In other words, a block of pixels representing motion of an image signal, or a motion block, is filtered with a low-pass filter (LPF) thereby eliminating the original definition to eliminate noise in the signal. On the other hand, a block of pixels representing no motion or a very small amount of motion, is not filtered.

The conventional interframe coding system carries a problem of quantization. Coding performance, according to the conventional system, is based upon a limited quantization which is designed optimally for a certain pattern of predictive error signal 15. In other words, the limited quantization can not deal effectively with coding signals of various patterns. When the encoder quantizes the signal with the limited quantization, the predictive error signal 15 is characterized with a poor coding efficiency and results in producing a poor coded error signal.

Ohtsuka et al: : "Development of 135 Mbit/s HDTV codec", Signal Processing Image Communication, vol. 4, no. 4/5, pages 379 - 387, XP 000 293 755 Amsterdam NL, discloses a coding system which includes an encoder for coding an input signal into a coded signal and a frame memory for outputting a predictive signal, which system comprises a coding controller responsive to the input signal and the predictive signal for calculating a characteristic (signal energy) of the input signal and for generating a coding control signal for controlling the encoder based on the characteristic. A comparison is made between characteristics of the interframe signal and the intraframe signal for determining the coding mode, namely interframe or intraframe coding.

Helgeson et al: "Impact of Recent Compression/Decompression Technologies in Video Distribution Systems", 11th Digital Avionics Systems Conference, 5 October 1992, pages 568 - 572, XP 000 366 739 Seattle, refers to the International consultative Committee for Telephone 'and Telegraph (CCITT) H. 261 document which specifies a method of communication for visual-telephony. CCITT is based on transform coding and temporal predictive coding with motion compensation. The CCITT H. 261 hybrid video coding algorithm utilises both intraframe and interframe coding schemes. The intraframe mode realises compression in the spatial dimension within a single frame which compression in the temporal mode is disabled. In addition to compression within a single frame, the interframe mode provides compression between consecutive frames in time. The CCITT encoder algorithm requires ad decoder to reconstruct and motion compensate the previous (N-1) frame and utilises this reconstructed image as a prediction for the current frame (N).

Finally, WO/88/10544 discloses a coding system for coding an input signal which comprises a local loop for outputting a predictive signal and a controller for receiving the difference of the input signal and the predictive signal and for generating a coding control signal for controlling the quantization of the error signal (difference of input signal and predictive signal). To this end, the mean value of the error signals in each block is calculated and a quantizer for each block is selected, using the calculate sigma value and the mean value of the error signals in each block, from a set of minimum square error two-variable quantizers.

The present invention is devoted to solve the above mentioned problem. An objective of this invention is to provide an interframe coding system and method which allow the encoder to code efficiently with coding signals of various patterns. This can lead to the overall coding efficiency of the interframe coding system and method, respectively.

This object according to the invention is solved for the system by the features of claim 1 and for the method by the features of claim 4. Preferred embodiments of the interframe coding system and method according to the invention are defined in the respective subclaims.

SUMMARY OF THE INVENTION

In Accordance with a first aspect of the invention, an interframe coding system for frame-to-frame coding of an input signal is provided, having a local decoding loop which includes an encode for interframe coding the input signal, decoding means for producing a decoded signal, and a frame memory which receives the decoded signal and outputs a predictive signal, the interframe coding system comprising a controller for receiving the input signal and the predictive signal from the frame memory, and for generating a coding control signal for controlling the encoder, characterized in that the controller includes a difference calculator for receiving the input signal and the predictive signal, for calculating the absolute values or squared values of pixel differences between said input signal and said predictive signal for each block of pixels and for outputting said values of pixel differences as a difference signal for each of said block of pixels; wherein a pixel difference calculator is provided for receiving said difference signal, for calculating a pixel difference signal based on said difference signal and the number of pixels per block and outputting said pixel difference signal to the encoder, and wherein the encoder includes means for quantizing the input signal with a step size and for changing the step size of quantization so that the quantizing error becomes smaller than said pixel difference signal.

In Accordance with a second aspect of the invention an interframe coding method for frame-to-frame coding of an-input signal, wherein the coding system includes. a local decoding loop including an encoder for interframe coding an input signal, decoding means for producing a decoded signal, and a frame memory for receiving the decoded signal and outputting a predictive signal, comprises the steps of: <ul id="ul0001" list-style="dash"><li>receiving the input signal and the predictive signal;</li><li>calculating the absolute values or squared values of pixel differences between said input signal and said predictive signal for each block of pixels;</li><li>outputting said values of pixel differences as a difference signal for each of said block of pixels;</li><li>calculating a pixel difference signal based on said difference signal and the number of pixels per block;</li><li>outputting said pixel difference signal to the encoder;</li><li>changing a step size of quantization in the encoder so that the quantizing error becomes smaller than said pixel difference signal; and</li><li>coding the input signal based on the step size of quantization.</li></ul>

BRIEF DESCRIPTION OF THE DRAWINGS

<dl id="dl0001" compact="compact"><dt>Fig. 1</dt><dd>shows a block diagram illustrating a configuration of an interframe coding system according to an embodiment of the present invention;</dd><dt>Fig. 2</dt><dd>shows a block diagram of a configuration of a filter controller according to the embodiment with reference to Fig. 1;</dd><dt>Fig. 3</dt><dd>is an explanation diagram of an operation of an encoder according to the embodiment with reference to Fig. 1;</dd><dt>Fig. 4</dt><dd>shows a block diagram of a configuration of an interframe coding system according to another embodiment of the present invention;</dd><dt>Fig. 5</dt><dd>shows a block diagram of a configuration of a controller according to the embodiment with reference to Fig. 4;</dd><dt>Fig. 6</dt><dd>is an explanation diagram of frame and field; and</dd><dt>Fig. 7</dt><dd>shows a block diagram of a conventional configuration of an interframe coding system.</dd></dl>

Embodiment 1.

Fig. 1 shows a configuration of the interframe coding system in accordance with the invention. The filter controller 21 outputs a pixel difference signal 25 or a control signal to the encoder 4 as well as the filter control signal 23 to the adaptive filter 22. The pixel difference signal 25 is the difference per pixel which is calculated by dividing the value of the difference signal 32 by the number of pixels in a block of pixels (e.g. 8x8=64) in the filter controller 21.

Fig. 2 shows a configuration of the filter controller 21 according to this embodiment which includes a difference calculator 30, a decision unit 31 and a pixel difference calculator 33. The difference calculator 30 basically calculates a differnce between the input image singal 12 and the motion compensation predictive signal 14 to output the difference signal 32. Specifically, the difference signal 32 is output in the following process: the difference calculator 30 calculates an absolute difference or squared difference between the signals for each pixel. Then the absolute values or squared values of pixel differnces are calculated for the processing unit or a block of 8x8 (64) pixels and output as the difference signal 32.

The value of the difference signal 32 is divided in the pixel difference calculator 33 by the number of pixels in a block of pixels (8x8=64) to produce the average value of the difference signal 32 in order to output a pixel difference signal 25. The pixel difference signal 25 is the threshold value of quantization which is output to the encoder 4. The encoder 4 decides the optimal quantization step size for coding the image signal under the condition that qantization error does not exceed the value of the pixel difference signal 25.

Fig. 3 shows the pixel difference of an image signal illustrating the operation of deciding the quantization step size in the encoder 4 in accordance with this embodiment.

In Fig. 3(a), D1 designates the pixel difference of a block containing eight pixels, D2 designates the pixel difference of another block of eight pixels, and D3 designates the pixel difference of another block of eight pixels. Those pixel differences are supposed to be given the following condition in the filter controller 21 as shown in Fig. 3(b): D1&gt;D2&gt;D3.

When the pixel difference signal 25 is input into the encoder 4, the encoder 4 selects the optimal quantization step size for an input image signal based upon the pixel difference signal 25, so that the value of quantization error should be less than the value of the pixel difference signal 25. With D1 as the pixel difference signal 25, for instance, the encoder 4 selects a quantization step size SS1 (SS1&lt;D1), with D2 as the signal a step size SS2 (SS2&lt;D2) is selected; and a step size SS3 (SS3&lt;D3) is selected for D3.

Under that condition, the encoder 4 quantizes an input image signal by selecting the optimal quantization step sizes based upon the value on the pixel difference signal 25 in the following relations as shown in Fig. 3(d): SS1&gt;SS2&gt;SS3.

Thus, the quantization step size becomes larger when the difference between the input image signal 12 and the motion compensation predictive signal 14 is large. On the other hand, the quantization step size becomes smaller when the difference between the two signals is smaller. Consequently this leads to a highly efficient coding performance minimizing the quantization error to have a high quality decoded image.

The pixel difference signal 25 is not the only signal used to decided the quantization step size in the encoder 4. The quantization step size is also controlled or limited by the size of an output buffer which temporarily stores coded signals to be output from the encoder 4. The size of the buffer can limit the quantity of coding signals. In this case, the quantization step size is limited by the size of the output buffer.

Embodiment 2.

With further reference to Fig. 1, the interframe coding system in accordance with the invention can provide a controller 21a to output a control signal 25 to the encoder 4 for controlling the quantization step size. In this embodiment the adaptive filter 22 and the filter control signal 23 are not necessarily provided as shown in Fig. 4.

Fig. 5 shows a configuration of the controller 21a according to this embodiment. Fig. 5 comprises Fig. 2, as modified by removing the decision unit 31. The operation of the difference calculator 30 and the pixel difference calculator 33 corresponds to that described in reference to Fig. 2.

With further reference to the embodiments hereinbefore, the frame memory 1 can store an image signal not by the frame but by the field as shown in Fig. 6. Fig. 6 shows the relation between the frame and field of an image signal. Fig. 6(a) illustrates that the frame of an image signal is consisted of two fields of an image signal, the first and second fields. Fig. 6(b) illustrates a frame composed of the first and second fields of an image signal by using an interlace mode.

Further, the filtering of an image signal can also be processed by the field.

With further reference to the embodiments hereinbefore, the frame memory 1 can store a multiple number of time-varying frames or fields including the past, present, and forthcoming frames or fields.

A distinctive feature of the interframe coding system of the present invention is that the coding controller controls the quantization step size based upon one of the three signals which are a difference signal between the input image signal and the signal from the frame memory, the input image signal, and the signal from the frame memory. This can contribute to reducing the amount of coding error in a coded image signal.

As stated hereinbefore, the interframe coding system and method of the present invention have the following advantageous effects.

An advantageous feature is that filtering can produce a highly coding efficient image signal by eliminating higher frequency components in the signal in the event that the predictive signal still contains a large amount of quantization error.

Another advantageous feature is that a pixel difference is used to control coding error, so that a decoded image can carry less coded error. This can contribute to a highly efficient prediction for the forthcoming frames of an image signal.