Predictive video coding system

0001The present invention relates to a coding system for coding a signal with high efficiency.

0002As is known in the art, means for eliminating redundant components included in an image signal is used for coding an image signal. A typical approach to image coding is the transform coding method wherein an image is divided into blocks, an orthogonal transform is carried out for each of the blocks, and the transform coefficients are encoded.

0003In the case of television signals such as an NTSC signal, interlaced scanning is used whereby an image signal of one frame is scanned twice, once in the odd field and once in the even field. The two fields scan different but complementary spaces of an image. The fields have image information at different times but there is a relatively strong correlation therebetween because the scanned lines of the two fields are alternate and adjacent. There is a technique in which coding is carried out after combining the fields and dividing them into blocks when coding an image signal produced by the interlaced scanning.

0004Fig. 1 is a block diagram of a conventional interframe predictive coding system described, for example, in the transactions on the 3rd HDTV International Work Shop, "A Study on HDTV Signal Coding with Motion Adaptive Noise Reduction" (Vol 3, 1989). In Fig. 1, this system comprises a frame memory 21, a motion detecting section 22, a subtracter 23, a coding section 24, a local decoding section 25, an adder 26 and a multiplexing section 27. Although omitted in this figure, the encoded data is decoded at a receiving side in order to reproduce the transmitted signal.

0005In operation, the motion of an object between the current field and the field of the same type of the preceding frame is detected block by block, the block consisting of a plurality of pixels of an input image signal 201 which is provided by the interlaced scanning method and formed of frames, each frame having both odd and even fields. The motion between odd fields is detected in the motion detecting section 22 by searching the block which has the most distinctive resemblance to the currently processing block among the already encoded blocks 202, adjacent to the position corresponding to the currently processing block in the odd fields stored within the frame memory 21. The degree of resemblance is evaluated by using an absolute sum of differential values or a square sum of differential values of the corresponding pixels in both blocks. The amount of motion in both horizontal and vertical directions between the current block and the block determined to be the most similar is provided as a motion vector 203. The frame memory 21 outputs a motion compensated prediction signal 204 corresponding to this motion vector 203.

0006A prediction error signal 205 obtained in the subtracter 23 by subtracting the motion compensated prediction signal 204 from the input signal 201 is applied to the coding circuit 24 in which the spatial redundancy is removed. Since low frequency components of an image signal generally occupy a greater part of the power thereof, information can be compressed by quantizing high power portions with a large number of bits and quantizing low power portions with a small number of bits. According to an example of this information compression method, the frequency conversion is carried out for an 8 x 8 pixels block by conducting an orthogonal transform such as a discrete cosine transform to scalar-quantize the transform coefficients. The scalar-quantized coding data 206 is sent to the local decoding section 25 and to the multiplexing section 27. The multiplexing section 27 conducts multiplexing and encoding for the coding data 206 and the motion vector 203 to output these signals to a transmission line 209.

0007Meanwhile, the local decoding circuit 25 executes the inverse operation of the operation in the coding section 24, namely the inverse scalar quantization and inverse orthogonal transform to obtain a decoded error signal 207. The motion compensated prediction signal 204 is added to the decoded error signal 207 in the adder 26 and stored in the frame memory 21 to detect motion of the odd field of the next frame.

0008In addition, the motion of the even fields of the input image signal 201 with respect to the already encoded field of the frame memory 21 is also detected for the coding of the motion compensated prediction error signal. As described above, in the conventional interframe predictive coding system, redundancy with respect to time included in moving image signals is removed by the motion compensated prediction coding and redundancy with respect to space is removed by the orthogonal transform.

0009Since the conventional interframe predictive coding system is structured to individually encode both the odd field and even field by predicting the current (present) odd field from the odd field of the already encoded frame and predicting the current even field from the even field of the already encoded frame, the encoding efficiency is low because the spatial correlation existing between the continuous fields, produced by the interlaced scanning method, is not used.

0010US-A-4 546 386 already discloses an adaptive predictive coding system for television signals, which comprise: <ul id="ul0001" list-style="none" compact="compact"><li>a storage section having a capacity large enough to store input television signals of at least one frame at all times;</li><li>a blocking section for producing from the television signals a block to be coded including a plurality of picture elements;</li><li>a movement detecting section for setting a standard block of the same shape as the block in a frame immediately preceding a current field of the block at a position spatially corresponding to the block, for reading out of the storage section picture elements included in each of reference blocks obtained by moving the standard block by at least one picture element and in the standard block and for selecting one of the reference blocks and the standard block as a most approximate block each picture element of which is most approximate in luminance to each picture element included in the block of the current field;</li><li>an intra-field predicting section for producing a prediction value of each picture element included in the block to be coded through using picture elements present in the current field;</li><li>an inter-field predicting section for producing a prediction value of each picture element included in the block to be coded through using picture elements present in the current field and the immediately preceding field;</li><li>an inter-frame predicting section for producing a prediction value of each picture element included in the block to be coded through using picture elements present in the current field and the immediately preceding field and included in the most approximate block;</li><li>a median selecting section for comparing the prediction values from the three predicting sections in terms of magnitude with one another and for selecting therefrom the prediction value assuming a median;</li><li>a prediction mode switching section for adopting, as the prediction value of each picture element included in the block to be coded, the prediction value from the inter-frame predicting section or the prediction value from the median selecting section depending on whether the most approximate block is the standard block or the reference block; and</li><li>a coding section for predictive-coding the block to be coded determined by the blocking section, through using the prediction value from the prediction mode switching section.</li></ul>

0011It is the object of the present invention to provide a method which enables encoding with higher prediction efficiency.

0012This object according to the invention is solved by a method comprising the features set out in claim 1. Preferred embodiments of the method according to the invention are defined in the subclaimes.

0013The invention refers to a method for producing a local decoded signal, wherein said local decoded signal is produced by adding a decoded error signal with a motion compensated predictive signal, and said decoded error signal is produced by decoding a coding data which is a coded predictive error signal developed from a difference between a first video image and a second video image of a motion video signal representative of sequential video images including first and second video images, which method is characterized by the steps: <ul id="ul0002" list-style="none" compact="compact"><li>storing the local decoded signal as plural odd and even image fields in a field memory;</li><li>generating plural predictive signals from said plural image fields stored in said field memory by operatively connecting said field memory;</li><li>generating an interpolated predictive signal which is different from any of the plurality of predictive signals by interpolating said plurality of predictive signals;</li></ul> wherein said motion compensated predictive signal is obtained from the plurality of predictive signals and the interpolated predictive signal.

0014With such an arrangement, the claimed method can provide stabilized prediction efficiency regardless of motion of an object by making reference to both fields of the already encoded frame for the purpose of prediction.

0015Since this method utilizes a predictive signal produced by interpolating the predictive signals from both fields of the already encoded frame, motion at the intermediate point of time and space of the two fields used for the prediction can be considered. Moreover this method also performs the function of a low-pass filter, whereby the prediction efficiency can be improved and the encoded image is stabilized.

0016The invention will be more fully understood from the following detailed description and the accompanying drawings in which: <dl id="dl0001"><dt>Fig. 1</dt><dd>is a block diagram showing the structure of a coding system of the prior art;</dd><dt>Fig. 2</dt><dd>is a block diagram showing the structure of an adaptive field/frame coding system of an embodiment of the present invention;</dd><dt>Fig. 3</dt><dd>is a diagram showing a exemplary input image signal;</dd><dt>Fig. 4</dt><dd>is a block diagram showing an example of the structure of an interpolating section shown in Fig. 2;</dd><dt>Fig. 5</dt><dd>is a diagram for explaining the operation of a motion detecting circuit;</dd><dt>Fig. 6</dt><dd>is a diagram for explaining the operation for using a motion compensated predictive signal in the embodiment shown in Fig. 2;</dd><dt>Fig. 7</dt><dd>is a block diagram showing another example of the interpolating section;</dd><dt>Fig. 8</dt><dd>is a block diagram showing an adaptive field/frame coding system according to embodiment of the present invention;</dd><dt>Fig. 9</dt><dd>is a block diagram showing an example of the structure of the blocking selection section;</dd><dt>Fig. 10</dt><dd>is a diagram showing a structural example of the block selected by the blocking selecting section;</dd><dt>Fig. 11</dt><dd>is a block diagram showing a structural example of the blocking forming section;</dd><dt>Fig. 12</dt><dd>is a block diagram showing a structural example of the blocking decomposing section;</dd><dt>Fig. 13</dt><dd>is a block diagram showing another structural example of the blocking selecting section;</dd><dt>Fig. 14</dt><dd>is a block diagram showing another structural example of the blocking selecting section;</dd><dt>Fig. 15</dt><dd>is a block diagram showing a structural example of the frequency analyzing section;</dd><dt>Fig. 16</dt><dd>is a diagram showing an example of the accumulated frequency components; and</dd><dt>Fig. 17</dt><dd>is a block diagram showing another structural example of the present invention.</dd></dl>

0017Referring to Fig. 2, a structural diagram of an adaptive field/frame coding system according to a first embodiment of the present invention is shown. The system includes an odd field memory 28 for storing local decoded signals of odd fields, and even field memory 29 for storing local decoded signals of even fields, and interpolation section 20 for interpolating a predictive signal with motion compensated from the two fields, and a selector 21 for selecting a predictive signal which gives the optimum prediction from three signals of the signals predicted from the odd and even fields and the interpolated predictive signal. In Fig. 2, sections 200, 300 and 500 enclosed by a broken line respectively denote motion detecting means, predicting error signal output means and coding means.

0018Fig. 3 shows a profile of input image signals 201 which are scanned by the interlaced scanning method, wherein the odd and even fields are alternately applied. Fig. 3 shows the fields in the coordinates where time is plotted on the horizontal axis and vertical direction on the vertical axis. In Fig. 3, K1 indicates an odd field of the first frame, while G1, an even field of the first frame. In the same manner, K2 is an odd field of the second frame, while G2, an even field of the second frame.

0019Fig. 4 is a block diagram of an example of the interpolating section 20. A simple arithmetic mean of the motion compensated prediction signal 204a from the inputted odd fields and the motion compensated prediction signal 204b from the inputted even fields is obtained and is used as an interpolation predictive signal 204c.

0020The operation will be explained with reference to Figs. 2, 3 and 4. Motion of the odd fields and even fields of the present frame in relation to the preceding frame is detected in units of blocks including pixels (n x m) in response to the input image signal 201 which is scanned by the interlace scanning method and includes the odd and even fields alternately. The motion of the odd fields between the present and the preceding frames is detected by searching, in the motion detecting section 22, the block which most resembles the currently processed block in the image signal 201 from the blocks adjacent 202a to the position corresponding to the currently encoded object in the already encoded odd fields stored within the odd field memory 28.

0021As shown in Fig. 5, for example, it is assumed that image H1 exists within one block unit (n x m) in the preceding frame, and the image moves to position H2 from position H1 in the present input image signal. The motion detecting section 22 outputs a motion vector 203 which indicates the block has moved horizontally to H2 from H1. In this case, since motion is not detected in the vertical direction, the motion vector 203 has the value of 0 with regard to vertical direction. The motion in the horizontal and vertical directions thus obtained is outputted as the motion vector 203.

0022The odd field memory 28 outputs a motion compensated prediction signal 204a corresponding to this motion vector 203. Similarly, compensation for motion of the even fields in the preceding frame is carried out in the motion detecting section 22, by searching the block resembling the currently processed block from the adjacent blocks 202b within the even field memory 29 and outputting the result as the motion vector 203. The motion compensated prediction signal 204b corresponding to this motion vector 203 is outputted from the even field memory 29.

0023The interpolation processing is carried out in the interpolating section 20 shown in Fig. 4, by using the motion compensated prediction signals 204a and 204b to generate the interpolation predictive signal 204c, signal 204a being generated by motion compensated in accordance with the motion vector 203 and provided from the odd field memory 28, and motion compensated predictive signal 204b being generated by motion compensated in accordance with the motion vector 203 and provided from the second field memory 9. A predictive signal having the minimum error signal power with respect to the currently encoding object block of the input image signal 201 is selected by the selector 21 from among the motion compensated prediction signal 204a obtained from the odd field, the motion compensated prediction signal 204b obtained from the even field, and the interpolated motion compensated prediction signal 204c, and then the predictive signal 210 is produced.

0024Fig. 6 is a diagram showing the operation explained above. It is assumed that the odd field memory 28 shown in Fig. 2 stores an odd field K1 of the preceding (previous) frame, while the even field memory 29 of Fig. 2 stores an even field G1 of the preceding frame. Here, the case where an odd field K2 and an even field G2 are included in the current (present) frame of the input image signal 201 will be discussed. First, when the odd field K2 is inputted, the motion compensated prediction signal 204a from the odd field K1 of the preceding frame stored in the odd field memory 28 is provided to the selector 21. In the same manner, the even field G1 of the preceding frame stored in the even field memory 29 is provided to the selector 21 as the motion compensated prediction signal 204b. Then, the data of K1 and G1 are applied to the interpolating section 20 and the interpolation processing as shown in Fig. 4 is conducted. Thereafter, such data is supplied to the selector 21 as the motion compensated prediction signal 204c. The selector 21 compares these three kinds of motion compensated prediction signals 204a, 204b, 204c and the input image signal 201 to select the prediction signal which has the minimum error signal power.

0025In the same manner, the selector 21 is responsive to the even field G2 of the current frame to receive the prediction signal 204a based on the odd field K1 stored in the odd field memory 28, the motion compensated prediction signal 204b based on the even field G1 stored in the even field memory 29, and the motion compensated prediction signal 204c obtained by the interpolation process on the basis of these motion compensated prediction signals 204a, 204b based on both fields, and to select the prediction signal which has the minimum error signal power.

0026Further, in the embodiment shown in Fig. 2, the simple arithmetic mean has been used for the interpolation section, but coding ensuring higher prediction efficiency can be realized by utilizing a weighted arithmetic mean taking into consideration field distance, as will be explained hereunder with reference to Fig. 7.

0027Fig. 7 is a block diagram of an example of the interpolation circuit 20. The motion compensated prediction signal 204a from the odd field is multiplied by a weight α based on the distance to the field to be encoded, and the motion compensated prediction signal 204b from the even field is multiplied by a weight β based on the distance to the field to be encoded. Thereafter, the arithmetic mean of these values is obtained and the output thereof is used as interpolation predictive signal 204c.

0028The practical value of the weighting by the interpolation section 20 in relation to the embodiment shown in Fig. 7 will be explained with reference to Fig. 6.

0029As shown in Fig. 6, when T is considered a unit of time for inputting an odd field or an even field, there is a time difference of 2T between odd field K1 and odd field K2. On the otherhand, there is a time difference of T between even field G1 and odd field K2. Thus, the weights α and β can be determined by utilizing such time differences. For example, since the odd field K1 has a time distance of 2T, the weight α is set to 1. Also, since even field G1 has a time distance of T from odd field K2, the value of weight can be increased for the field having the lesser time distance by setting the value of β to 2. In the same manner, odd field K1 has a time distance of 3T from even field G2 and even field G1 has a time difference of 2T. Thus, ist is possible to give the value of weight which is proportional to the time difference by setting α to 2 and β to 3 for weighting even field G2.

0030In the embodiment shown in Fig. 7, the weights α and β are determined in the interpolating section on the basis of time distance. However, it is also possible that the weight α to be given to the odd field is always set, for example, larger or smaller than weight β to be given to the even field regardless of the time distance. Further, in this embodiment, weights α and β used for the odd fields are different from those used for the even fields, but the weights for the odd fields may be equal to those for the even fields. In addition, in this embodiment, only weights α and β are used, but the weights may be determined in accordance with the other coefficients, for example, a coefficient having a quadratic function or another function having particular characteristics. Moreover, weights α and β do not have to be restricted only to one kind of value; it is possible that several kinds of weights α and β are prepared and selected in accordance with the kind of input signal or the characteristic of input signal.

0031Another embodiment of the present invention will be explained with reference to Fig. 8.

0032The embodiment shown in Fig. 8 comprises a blocking selection section 82 for selecting between an individual blocking of a prediction error signal for the odd and even fields and a non-interlace blocking including both odd and even fields; a blocking forming section 83 for conducting the blocking in accordance with the output of the blocking selection section 82; and a blocking decomposing section 84 for decomposing the blocking to form the original field in accordance with the block selection output. Section 400 enclosed by a broken line denotes blocking means and the other sections 200, 300, 500 are similar to those shown in Fig. 2.

0033Fig. 9 is a block diagram of an example of the blocking selection section 82. The prediction error signal 205 is stored in the oddd field memory 31 for the odd field and in the evend field memory 32 for the even field. As shown in Fig. 10(a) and 10(b), a block of p = 16, q = 16 is considered. The individual field blocking section 33 executes the blocking including the pixels of either of the odd or even field within the block of (p pixels x q lines), and these pixels are encoded in a coding section 35. As shown in Fig. 10(c), a non-interlace blocking section 34 executes the blocking of (p pixels x q lines) included in the block by alternately arranging the pixels of both odd and even fields, and these pixels are encoded in a coding circuit 36. The information quantity comparing section 37 compares the quantity of data encoded in the coding section 35 and the coding circuit 36, and outputs a blocking selection signal 211 indicating the blocking having the least amount of information.

0034Fig. 11 is a block diagram of an example of the blocking forming section 83. The prediction error signal 205 is stored in the odd field memory 41 for the odd field and in the even field memory 42 for the even field. In accordance with the blocking selection signal 211 supplied from the blocking selection sectidon 82, the blocking forming section 43 selects the blocking of the prediction error signals stored in the odd field memory 41 and even field memory 42 from the blocking including pixels of either of the odd or even field within the block of (p pixels x q lines) and the blocking including pixels of both odd and even fields within the block of (p pixels x q lines), and then outputs the blocked prediction error signal.

0035Fig. 12 is a block diagram of an example of the blocking decomposing section 84. The data decoded by a local decoding circuit 25 is applied to the blokking decomposing section 44 in which the blocking is decomposed in accordance with the blocking selection signal 211 from the blocking selection section 82, and the decomposed block is then stored in the individual field memories 45, 46. The stored data is supplied as a decoded error signal 207.

0036The operation of this embodiment is explained hereunder.

0037The prediction error signal 205 obtained by subtracting the prediction signal 210 from an input signal 201 in a difference circuit 23 is sent to the blocking forming section 83 shown in Fig. 11 and to the blocking selection section 82 shown in Fig. 9. The blocking selection section 82 produces the blocking selection signal 211 for selecting the blocking including the pixels of either the odd or even field in the block of (p pixels x q lines), or the blocking including the pixels of both odd and even fields in the block of (p pixels x q lines). The blocking forming section 83 conducts individual field blocking or non-interlace blocking in units of (p x q) blocks in accordance with the blocking selection signal 211.

0038The blocked signal is applied to the coding circuit 24. The coding section 24 execute the orthogonal transform and sends the encoded data 206 which is a scalar-quantized transform coefficient to both the local decoding section 25 and the multiplexing section 28.

0039After the inverse scalar-quantization and inverse orthogonal transform by the local decoding section 25, the data is decomposed into the odd and even fields in the blocking decomposing section shown in Fig. 12 which decomposes the blocking into the fields in accordance with the blocking selection signal 211 in order to obtain the decoded difference signal 207. The local decoded signal 208 obtained by adding a predictive signal 210 to the decoded difference siganl 207 in the adder 207 is stored in the first field memory 28 when it is the odd field or in the second field memory 29 when it is the even field, to detect the motion of each field of the next frame.

0040In this embodiment, a unit of blocks is formed of p = 16, q =16, but it is desirable that the values of p and q have the following relationship with the block size n x m used by the motion detecting section 22 as explained in the embodiment shown in Fig. 2:<maths id="math0001"><math display="block"><mrow><mtext>p = n, q = 2m.</mtext></mrow></math><img file="EP0984635B1_D0001.tif" /></maths>

0041Since DCT transform is often carried out in the block unit of 8 pixels x 8 lines, the size of 16 pixels x 16 lines combining four block units is selected as the values of p and q in the blocking forming section. In this example, since p = n, n = 16 pixels. Also, since q = 2m, m = 8. Thus, it is desirable that the number of lines be reduced to 8 because the motion detecting section 22 detects motion for both the odd and even fields. Meanwhile, since it is possible to employ the blocking combining the odd field and even field in the blocking forming section, it is desirable to form a block fo 16 lines including the odd and even fields.

0042In the embodiment shown in Fig. 8, the blocking has been selected by comparing the quantity of information generated as shown in Fig. 9, but coding based on the quality of encoding can be realized by selecting the blocking on the basis of the comparison of encoding quality as shown in Fig. 13.

0043Fig. 13 is a block diagram of an example of the blocking selection section 82. The predicting error signal 205 is stored in the odd field memory 51 for the odd field and in the even field memory 52 for the even field. The individual field blocking section 53 realizes the blocking including the pixels of either the odd field or the even field within the block of (p pixels x q lines), and the coding/decoding section 55 enables encoding/decoding. At the same time, the non-interlace blocking section 54 realizes the blocking including the pixels of both fields within the block of (p pixels x q lines), and the coding/decoding circuit 56 enables coding/decoding. The difference between the encoded/decoded data of the individual field blocking and the data just before the encoding is compared with the difference between the encoded/decoded data of the combined field blocking and the data just before the encoding, by the error comparator 59 in order to select the blocking with less errors and to provide an output as the blocking selection signal 211.

0044In the embodiment shown in Fig. 8, the quantity of generated information has been compared for the selection of the block, while in the embodiment shown in Fig. 13, the encoding errors have been compared. However, encoding with higher efficiency can be realized when conducting encoding utilizing the orthogonal transform, by selecting the blocking on the basis of the comparison of frequency components produced by the difference of blocking as shown in Fig. 14.

0045Fig. 14 is a block diagram of an example of the blocking selection circuit 82. The predicting error signal 205 is stored in the odd field memory 61 for the odd field and in the even field memory 62 for the even field. The individual field blocking section 63 executes the blocking including the pixels of only either the odd field or even field within the block of (p pixels x q lines), and a frequency analyzing section 65 such as that shown in Fig. 15 executes the frequency analysis. The non-interlace blocking circuit 64 executes the blocking including pixels of both fields within the block of (p pixels x q lines), and a frequency analyzing circuit 66 such as that shown in Fig. 15 executes the frequency analysis. The blocking with fewer high-frequency components is selected from the individual field blocking and the combined field blocking to output the blocking selection signal 211.

0046Fig. 15 is a block diagram of an example of the frequency analyzing sections 65 and 66. The signal obtained by individually blocking the odd and even fields from the individual field blocking circuit 63, and the signal obtained by blocking the pixels of both odd and even fields from the non-interlace blocking section 64, are supplied to sections 65 and 66. These signals are converted to a signal in the frequency domain from a signal in the pixel domain using the orthogonal transform 68. The high-frequency components are extracted from the converted signal in the frequency domain by a high-frequency component selector 69 and the extracted high-frequency components are totaled by a high-frequency component accumulator 70. The accumulated high-frequency components are compared in a high-frequency component comparing section 67 to select the blocking with fewer amount high-frequency components.

0047Fig. 16 shows an example of the components accumulated by the high-frequency component adder 70 from the orthogonal transformed frequency domain signal. Here, eight components, for example, having the maximum frequency component in the vertical frequency component, are selected.

0048In this embodiment, the coding section 24 does not use the selection information of predictive signals or the selection information of blocking, but according to another embodiment shown in Fig. 17, finer control is possible and high encoding quality can be realized by inputting an output of the selector 11 as the selection signal for the predictive signal and the blocking selection signal as the selection signal for the blocking to the coding section 24 and by controlling the encoding characteristic with the selected prediction signal and the information of the selected blocking.

0049As explained above, the embodiment of Fig. 2 relates to a system for realizing predictive coding of an input image signal obtained by the interlaced scanning method with the motion compensation. The system includes motion detecting means for obtaining, for the odd or even field of the input image signal, the amount of displacement, in order to carry out the individual motion compensated prediction, in units of the block of (n pixels x m lines) (n and m: positive integer) from both the odd and even fields of the already encoded frame, and the prediction error signal output means for selecting, with a selector 21, the predictive signal indicating the optimum prediction from signals including a first predictive signal 204a obtained by the motion compensation from the odd field, a second predictive signal 204d obtained by the motion compensation from the even field, and a third predictive signal 204c obtained by interpolating the first and second predictive signals in order to obtain the difference from the field of the input signal and output the result as the prediction error signal.

0050Moreover, the embodiment of Fig. 2 is an adaptive field/frame coding system characterized in that the interpolation means for obtaining the third predictive signal is the simple arithmetic mean of the first predictive signal and the second predictive signal.

0051Thus, the hardware can be minimized in size and encoding with higher prediction efficiency can be realized by generating an interpolation signal of the predictive signal by simply obtaining the arithmetic mean of both predicted odd and even fields with motion compensation.

0052Further, the embodiment of Fig. 7 is an adaptive field/frame coding system characterized in that the interpolation means for obtaining the third predictive signal is the weighted arithmetic mean of the first predictive signal and the second predictive signal, also considering the time distance of the field used for the prediction and the field to be encoded.

0053Thus, encoding ensuring very high prediction efficiency can be realized by generating the interpolation signal from the weighted arithmetic mean of both predicted odd and even fields with the motion compensation, while considering the time distance of the field used for the prediction and the field to be encoded.

0054The embodiment shown in Fig. 8 is an adaptive field/frame coding system comprising means for enabling encoding by selecting blocking including the pixels of either the odd field or even field within the block of (p pixels x q lines), or blocking including the pixels of both odd and even fields within the block of (p pixels x q lines), in order to encode the prediction error signal for the odd and even fields of the input image signal in units of the block of (p pixels x q lines) (p and q: positive integer).

0055Moreover, the embodiment shown in Fig. 8 is an adaptive field/frame coding system characterized in that the blocking means for enabling encoding while selecting the blocks comprises selecting means for selecting the blocking with less information for encoding from blocking including the pixels of only one of the odd field and even field within the block of (p pixels x q lines), and blocking including the pixels of both odd and even fields within the block of (p pixels x q lines).

0056The embodiment shown in Fig. 13 is an adaptive field/frame coding system characterized in that the blocking means for enabling encoding while selecting the blocks comprises means for selecting the blocking with less encoding error from blocking including the pixels of only one of the odd field and even field within the block of (p pixels x q lines), and blocking including the pixels of both odd and even fields within the block of (p pixels x q lines).

0057The embodiment shown in Fig. 14 is an adaptive field/frame coding system characterized in that the blocking means for enabling encoding while selecting the blocks comprises selecting means for selecting the blocking with less high-frequency components included in the signal to be encoded from blocking including the pixels of only one of the odd field and even field within the block of (p pixels x q lines), and blocking including the pixels of both odd and even fields within the block of (p pixels x q lines).

0058In addition, the embodiment shown in Fig. 17 is an adaptive field/frame coding system characterized by enabling encoding while selecting the quantization characteristic of the transform coefficient in accordance with the selected predictive signal and the selected blocking, in the case of employing the orthogonal transformer and carrying out encoding by the quantization of transform ceofficient in the coding section for the encoding in units of the block of (p pixels x q lines).

0059In the above embodiments, an input image signal 201 is formed of the frame including the odd field and even field. However, the use of the odd field and even field is intended to show only an example, and the field is not restricted to the odd or even field. The present invention can be useful whenever one frame is divided into fields, the odd field and even field being only examples of such fields of a frame. For instance, the present invention can also be applied to a case of storing data by dividing the frame into two fields every two lines by, for example defining the first field as the 1st and 2nd lines and the second field as the 3rd and 4th lines, and defining the first field as the 5th and 6th lines and the second field as the 7th line and 8th line, etc. Moreover, in addition to dividing a frame into two kinds of fields, such as the odd field and the even field or the first field and the second field, the present invention can also be applied to the case of dividing a frame into more than two fields, for example, three or four kinds of fields. In such a case, the number of field memories corresponds to the number of kinds of fields, and the processing explained above is carried out for each field.

0060In the above embodiments, the blocking selection section selects the blocking from two kinds of blocking, including the blocking of the pixels of only one of the odd field and even field and the blocking of the pixels of both odd and even fields. However, the blocking may include various combinations when two or more fields are prepared in addition to the odd and even fields. The blocks shown in Figs. 10(a), (b), (c) are only examples and various block forming methods may be used to form the block other than the blocks of Fig. 10.

0061In the above embodiments, the blocking means shown in Fig. 8 is used with the prediction error signal output means and motion detecting means. Even if the sections other than the blocking means 400 are replaced with conventional means, the 3rd and 4th aspects explained above can be provided.

0062According to the 1st and 2nd aspect explained above, a stable encoded image with high efficiency can be obtained by individually searching the motion from each field of the already.encoded frame to predict each field and by conducting adaptive prediction from the searched motion compensated predictive signals (and interpolation signals).

0063In addition, according to the 3rd and 4th aspects explained aboved, a stabled encoded image with high efficiency can also be obtained by adaptively selecting the encoding from the blocking of the pixels of only one of the fields of the frame to be encoded, and the encoding after conducting the blocking of the pixels of the respective fields when encoding the prediction error signal.