Coding decoding method and device for motion picture

Description: TECHNICAL FIELD [Detailed description of the invention]

【0001】

[Industrial application field]

The present invention relates to a moving image code decoding method and an apparatus.

【0002】

[Conventional technology]

In the conventional technique, it is common to perform code decoding of a moving image by using motion compensation frame-to-frame prediction processing. As an example, in the moving image coding decoding method H.261 internationally standardized by CCITT, the input image is divided into blocks of 16 pixels and 16 lines, and motion compensation frame-to-frame prediction is executed for each block using the reference image. .. As the reference image, a decoded image obtained by decoding the last transmitted image-encoded data inside the coding apparatus is used. The auxiliary information used for the motion compensation frame-to-frame prediction and the error signal for the prediction are encoded and transmitted.

【0003】

In addition, the "image communication method" described in JP-A-62-068384 is characterized by the difference between the standard part data stored in the database in advance and the extracted part by extracting the element parts such as the face and arms that make up the subject from the input image. Encoded and transmitted as a parameter. On the decoding side, the plurality of standard partial data are transformed by the parameters and combined to synthesize a reproduced image.

【0004】

[Problems to be Solved by the Invention]

In the conventional motion compensation frame-to-frame prediction method H.261, the auxiliary information and the prediction error signal used for motion compensation are encoded and transmitted to the decoding side. Therefore, when the transmission capacity is small, there are problems that the coded data cannot be sufficiently transmitted and the decoded image is significantly deteriorated, or the frame rate of the decoded image is significantly lowered. Further, in the "image communication method" described in Japanese Patent Application Laid-Open No. 62-068384, only the extracted parameters are encoded and transmitted, so that it can be realized with a small transmission capacity, but a reproduced image is obtained by transforming the standard partial data prepared in advance. Therefore, there is a problem that the subtle features of the input image cannot be expressed and the image becomes artificial and unnatural.

【0005】

An object of the present invention is a case where the current input image is analyzed, compared with a plurality of input images code-decoded in the past to determine whether or not to encode the input image, and it is determined that the coding is not performed. Information for selecting the most similar past input image in the comparison is transmitted to the decoding side, and the decoding side decodes and displays the encoded data of the encoded input image and keeps a record. When the selection information is received, the corresponding coded data is selected from the retained coded data and decoded and displayed, so that an image similar to the input image can be decoded and reproduced with good image quality even when the transmission capacity is small. It is an object of the present invention to provide a code decoding method and an apparatus for a moving image.

【0006】

[Means for solving problems]

In the coding method of a moving image, the present invention extracts a feature amount of an input image as a parameter at the time of coding, records and holds the extracted parameter according to an instruction of a coding control signal, and records the above in the past. The plurality of parameters are output as reference parameters, the extracted parameters are compared with the plurality of reference parameters, the reference parameter having the highest similarity is selected, and the coding control signal is selected according to the value of the maximum similarity. Is output or the reference control signal is output to the decoding side, the input image is encoded according to the instruction of the coding control signal, output as encoded data to the decoding side, and at the time of decoding. Is to input either the coded data or the reference control signal for each input image from the coding side and record and hold the coded data or refer to the coded data recorded in the past. It is characterized in that one is selected according to an instruction of a control signal and output as reference data, and either the coded data or the reference data is decoded and output as an output image.

【0007】

According to the present invention, in a moving image coding / decoding device, the coding device records and holds a means for extracting a feature amount of an input image as a parameter and the extracted parameter according to an instruction of a coding control signal, and in the past. The means for outputting the plurality of recorded parameters as reference parameters and the reference parameter having the highest similarity by comparing the extracted parameters with the plurality of reference parameters are selected, and the above-mentioned is described according to the value of the maximum similarity. A means for either outputting a coding control signal or transmitting a reference control signal to the decoding side, and the input image is encoded according to the instruction of the coding control signal and sent to the decoding side as coded data. The decoding device includes means for transmitting, receives either the coded data or the reference control signal from the coding side for each input image, and records and holds the coded data or said in the past. A means for selecting one from a plurality of recorded coded data according to the instruction of the reference control signal and outputting it as reference data, and a means for decoding either the coded data or the reference data and outputting it as an output image. It is characterized by having.

【0008】

Further, in the method for decoding a moving image, the present invention extracts a feature amount of an input image as a parameter at the time of coding, records and holds the extracted parameter according to an instruction of a coding control signal, and has described the above in the past. A plurality of recorded parameters are output as reference parameters, the extracted parameters are compared with the plurality of reference parameters, the reference parameter having the highest similarity is selected, and the coding control is performed according to the value of the maximum similarity. Either the signal is output or the reference control signal is output to the decoding side, the input image is encoded according to the instruction of the coding control signal, output as encoded data to the decoding side, and the extraction is performed. The difference between the selected parameter and the selected reference parameter is output to the decoding side as a composite parameter, and at the time of decoding, either the coded data or the reference control signal is input from the coding side for each input image. Then, the coded data is recorded and held, or one of the held plurality of coded data is selected according to the instruction of the reference control signal and output as reference data, and either the coded data or the reference data is output. It is characterized in that the data is decoded and output as a decoded image, and the output image is synthesized and output from the decoded image by using the synthesis parameter.

【0009】

Further, according to the present invention, in a moving image coding / decoding device, the coding device records and holds a means for extracting a feature amount of an input image as a parameter and the extracted parameter according to an instruction of a coding control signal, and in the past. The means for outputting the plurality of recorded parameters as reference parameters and the reference parameter having the highest similarity by comparing the extracted parameters with the plurality of reference parameters are selected according to the value of the maximum similarity. A means for either outputting the coding control signal or transmitting the reference control signal to the decoding side, and the input image is encoded according to the instruction of the coding control signal and decoded as encoded data. A means for transmitting to the side and a means for outputting the difference between the extracted parameter and the selected reference parameter to the decoding side as a composite parameter are provided, and the decoding device is provided with the coded data or the coded data for each input image. One of the reference control signals is received from the coding side, and the coded data is recorded and held, or one of the held plurality of coded data is selected according to the instruction of the reference control signal and used as reference data. A means for outputting, a means for decoding either the coded data or the reference data and outputting as a decoded image, and a means for synthesizing and outputting an output image from the decoded image by using the synthesis parameter are provided. It is characterized by that.

【0010】

[Example]

An embodiment of the first moving image code decoding method and apparatus of the present invention will be described with reference to FIG.

【0011】

FIG. 1 is a block diagram of an embodiment of an apparatus that realizes the code decoding method of the first moving image of the present invention.

【0012】

The moving image code decoding device has a feature amount extraction circuit 12 that extracts the feature amount of the input image as a parameter, and records and holds the extracted parameter according to the instruction of the coding control signal, and the recorded parameter in the past. The parameter memory 14 that outputs the above as a reference parameter, the extracted parameter is compared with the plurality of reference parameters, the reference parameter having the highest similarity is selected, and the coding control signal is selected according to the value of the maximum similarity. The comparison circuit 13 that either outputs or transmits the reference control signal to the decoding side, and the input image is encoded according to the instruction of the coding control signal and transmitted as encoded data to the decoding side. Reference data is selected from a plurality of coded data recorded by recording and holding the received coded data or by instructing the reference control signal with a coding device composed of an image coding circuit 11. It is composed of a decoding device including a coded data memory 15 that outputs as an output image, and an image decoding circuit 16 that decodes the coded data or the reference data and outputs the output image.

【0013】

The operation of this embodiment will be described. Here, the operation of the code decoding process for the current input image 101 will be described on the premise that a plurality of images selected from the past series of input images 101 have already been code-decoded.

【0014】

On the coding side, the feature amount extraction circuit 12 first analyzes the input image 101 and extracts the parameter 103, which is the feature amount of the image. Examples of the extraction parameter 103 are shown in FIGS. 3 (a), (b), and (c), and each of them will be described.

【0015】

(A) in Fig. 3 is the time t_iInput image I (t_i) Extracts the subject area and its center coordinate position g_iAnd the inclination of the spindle θ_iIs an example in which and is used as the parameter. In this case, first, a closed region in the input image with little change in brightness or hue is extracted and regarded as one subject region. Next, the average value of the coordinate positions of each pixel constituting the subject area is calculated, and the g_iAnd. Further, the main axis of the subject area is calculated, and the inclination of the image with respect to the vertical axis is determined by θ._iAnd. In the example of (b) in Fig. 3, the time t_iInput image I (t_i) Is extracted from the subject area, and multiple feature points (p) on the contour line of the subject area are extracted._i, q_i, r_i, s_i) Is an example with the coordinate position as a parameter. In this case, the contour line of the subject area extracted from the input image by the above method is approximated by a polygon, and each vertex of the polygon is a feature point (p)._i, q_i, r_i, s_i). (C) in Fig. 3 is an example that can be used when the input image is known to be a face image in advance, and the time t_iInput image I (t_i) Extracts the entire face and the eyes and mouth, and the center coordinate position f of the face_iAnd a value (er) indicating the open / closed state of each eye and mouth_i, el_i, m_i) And are the extraction parameters. In this case, first, an elliptical region having a flesh-colored hue is extracted from the input image as a face portion, and the center coordinate position thereof is set to f._iAnd. Furthermore, in the face part, the f_iThe position of the eye or mouth is roughly predicted from the relative positional relationship with the above, and regions having different brightness and hue from the surroundings in the vicinity of the predicted position are extracted as the eye and mouth parts, respectively. The open / closed state of the eyes and mouth is the value of the vertical size of the extracted part (er)._i, el_i, m_i). The plurality of parameters described above may be any one or a combination of a plurality of parameters to form an extraction parameter 103.

【0016】

In the comparison circuit 13, the extraction parameter 103 is compared with the plurality of reference parameters 104 supplied by the parameter memory 14, and the reference parameter having the highest similarity is selected. In the selection, the absolute difference between the two parameters to be compared is used as the difference coefficient, and it is determined that the smaller the coefficient value, the higher the similarity. When the parameter is a set of a plurality of elements, the weighted addition value of the difference absolute value for each element is used as the difference coefficient. The weighting ratio may be uniform, or those having a high possibility of erroneous detection such as feature points of the outline of the outer shape of the subject are reduced, and those having a low possibility of erroneous detection such as the coordinate position of the center of the area are increased. Alternatively, when the input image is a face image and the extraction parameters consist of a plurality of elements of different types, the weighting of the elements related to the movement of the eyes and mouth is increased in order to more accurately detect the change in the facial expression. May be good.

【0017】

The comparison circuit 13 compares the difference coefficient in the selected parameter with a predetermined threshold value, and outputs either the reference control signal 106 or the coded control signal 105 according to the determination result. Specifically, when the difference coefficient is smaller than a predetermined threshold value, the data index corresponding to the selected parameter is transmitted to the decoding side as the reference selection signal 106. The data index also has a one-to-one correspondence with coded data 102, which encodes the original input image of the selected parameter. When the reference selection signal 106 is output, all the coding processing for the current input image 101 is completed. On the other hand, when the difference coefficient is larger than the predetermined threshold value, the reference selection signal 106 is not output, and the coding control signal 105 instructing the coding of the current input image 101 and the record keeping of the extraction parameters is output. To do. That is, if the current input image 101 is similar to a specific image that has been code-decoded in the past, only a simple signal instructing the decoding side to display the similar image is transmitted, and the similar image is transmitted. Only when is not found in the past coding history, the current input image 101 itself is encoded and transmitted to the decoding side. The threshold value may be set to a fixed value prior to a series of code decoding processes, or may be changed according to the severity of fluctuations in the extraction parameters. As a specific example, the parameters extracted from the input image of the previous frame are held with a delay inside the comparison circuit 13, and the absolute value dp of the difference between the frames between this and the extracted parameters in the current input image is used as the proportional coefficient. Formulate the threshold Th Th = (dp / dn) × Th (init) May be updated for each image frame. Here, dn and Th (init) are predetermined values prior to a series of code decoding processes.

【0018】

The image coding circuit 11 encodes the current input image 101 according to the instruction of the coding control signal 105 and transmits the coded data 102 to the decoding side. As the coding process, an image coding method defined as an international standard in ISO / JPEG can be used. Specifically, the input image 101 is divided into blocks of a predetermined size, discrete cosine transform, quantization and zigzag scan transform are performed for each block, and the coded data 102 is replaced with a variable length code. Further, when continuously encoding input images that are close in time, CCITT.261 or the like described above in the conventional technique can be used. Specifically, the last transmitted coded data is decoded by the local decoding device inside the image coding circuit 11 and stored in the memory, and the decoded image is used as a reference image for motion compensation frame-to-frame prediction code. Performs conversion on the current input image.

【0019】

The parameter memory 14 also additionally records the current extraction parameter 103 according to the instruction of the coding control signal 105. The record corresponds to recording and holding as history information that the current input image 101 is encoded and transmitted to the decoding side. The parameter memory 14 records and holds the parameters extracted for each of the plurality of input images that have been code-decoded in the past, and outputs these as a plurality of reference parameters 104. If the parameter memory 104 is full and it is difficult to add the parameter memory 104 during the additional recording, an appropriate parameter that has already been recorded is deleted, and the extraction parameter 103 is newly recorded. The parameter to be deleted is selected by using the history selected as the optimum reference parameter in the comparison circuit 13 by the method described above. Specifically, the comparison circuit 13 records the selected time for each reference parameter, and deletes the parameter having the oldest selected last time. Alternatively, the number of times selected may be accumulated for each reference parameter in a plurality of frame periods up to the present, and the parameter having the least number of selections may be deleted.

【0020】

On the decoding side, the coded data 102 received in the coded data memory 15 is recorded and held. Further, the received coded data 102 is directly supplied to the image coding circuit 16. On the other hand, when the reference selection signal 106 is received instead of the coded data 102, the coded data memory 15 indicates one of the plurality of coded data recorded and held by the reference selection signal 106. Select and output as reference data 107. When holding the record, if the coded data memory 15 is already full and additional recording is difficult, delete one of the plurality of coded data already recorded and renew it. The coded data 102 is additionally recorded. As the method for selecting the coded data to be deleted, the same method as the method for selecting the parameter to be deleted in the parameter memory 14 described above is used. Further, the maximum number of coded data that can be recorded and held is also the same as the maximum number of parameters that can be registered in the parameter memory 14 described above. With the above settings, the update of the record of the coded data memory 15 and the update of the record of the parameter memory 14 on the coding side can be automatically realized in synchronization with each other.

【0021】

The image decoding circuit 16 decodes any one of the supplied coded data 102 or reference data 107, and outputs an output image 108 for display reproduction.

【0022】

The series of code decoding processes described above can also be realized by the configuration example shown in FIG. The moving image code decoding device of FIG. 4 has a feature amount extraction circuit 42 that extracts the feature amount of the input image as a parameter, and records and holds the extracted parameters according to the instruction of the coding control signal, and has a plurality of past features. The parameter memory 44 that outputs the extracted parameter as a reference parameter in the input image and the reference parameter having the highest similarity by comparing the extracted parameter with the plurality of reference parameters are selected and set to the value of the maximum similarity. The comparison circuit 43, which either outputs the coding control signal or transmits the reference control signal to the decoding side, and the input image are encoded and encoded according to the instruction of the coding control signal. A coding device composed of an image coding circuit 41 to be transmitted to the decoding side as data and a series of received coded data are decoded to generate a decoded image, which is output as an output image for display reproduction. The image decoding circuit 46 to record and hold the decoded image, and when the coded data is not received, select one from the held plurality of decoded images according to the instruction of the reference control signal and output the output image. It is composed of an image memory 45 that outputs as.

【0023】

In the example of FIG. 4, on the coding side, the coded data 402 of the input image and the reference control signal 406 are output in the same manner as described above with reference to FIG. On the decoding side, the image decoding circuit 46 decodes the coded data 402 and outputs the decoded image 407. At the same time, the decoded image 407 is also output as an output image 408 for display reproduction.

【0024】

The image memory 45 records and holds the decoded image 407. When the reference control signal 406 is received instead of the coded data 402, the image memory 45 selects and displays one of the plurality of decoded images recorded and held by the reference selection signal 406. Output as output image 408 for playback. When holding the record, if the image memory 45 is already full and additional recording is difficult, one of the plurality of decoded images already recorded is deleted, and the decoded image is newly recorded. Record additional 407. The method of selecting the decoded image to be deleted uses the same method as the method of selecting the parameter to be deleted in the parameter memory 44. Further, the maximum number of decoded images that can be recorded and retained is also the same as the maximum number of parameters that can be registered in the parameter memory 44 described above. With the above settings, the update of the record of the image memory 45 and the update of the record of the parameter memory 44 on the coding side can be automatically realized in synchronization with each other.

【0025】

Next, an example of the code decoding method and the apparatus of the second moving image of the present invention will be described with reference to FIG. FIG. 2 is a block diagram of an embodiment of an apparatus that realizes the code decoding method for moving images of the present invention.

【0026】

The moving image code decoding device includes a feature amount extraction circuit 22 that extracts the feature amount of the input image as a parameter, and a plurality of the extracted parameters that are recorded and held according to the instruction of the coding control signal and recorded in the past. The parameter memory 24 that outputs the parameter of the above as a reference parameter, the extracted parameter is compared with the plurality of reference parameters, the reference parameter having the highest similarity is selected, and the encoding is performed according to the value of the maximum similarity. A comparison circuit 23 that either outputs a control signal or transmits a reference control signal to the decoding side, and the input image is encoded according to the instruction of the coding control signal and sent to the decoding side as encoded data. A coding device including an image coding circuit 21 to be transmitted, a diffifier 25 for obtaining a difference between the extracted parameter and the selected reference parameter and transmitting the difference as a synthesis parameter to the decoding side, and the received code. The coded data memory 26 that records and holds the coded data or selects one from the plurality of coded data recorded by the instruction of the reference control signal and outputs the coded data as reference data, and the coded data or the reference data. From a decoding device composed of an image decoding circuit 27 that decodes any of them and outputs it as a decoded image, and an image synthesis circuit 28 that deforms and synthesizes the decoded image using the synthesis parameter and outputs an output image. Become.

【0027】

The operation of this embodiment will be described. Here, the operation of the code decoding process for the current input image 201 will be described on the premise that a plurality of images selected from the past series of input images 201 have already undergone the code decoding process.

【0028】

On the coding side, first, the feature amount extraction circuit 22 analyzes the input image 201, and extracts the parameter 203, which is the feature amount of the image. As the extraction parameter 203, the parameters described above in the examples (a), (b), and (c) of FIG. 3 are used in combination.

【0029】

In the comparison circuit 23, the extraction parameter 203 is compared with a plurality of reference parameters 204 supplied by the parameter memory 24, and the reference parameter having the highest similarity is selected. Further, the comparison circuit 23 compares the difference coefficient between the extraction parameter 203 and the selected parameter with a predetermined threshold value, and outputs either the reference control signal 206 or the coding control signal 205 according to the determination result. Specifically, when the difference coefficient is smaller than a predetermined threshold value, the data index corresponding to the selected parameter is transmitted to the decoding side as a reference selection signal 206. The data index also has a one-to-one correspondence with coded data 202, which encodes the original input image of the selected parameter. On the other hand, when the difference coefficient is larger than a predetermined threshold value, the reference selection signal 206 is not output, and the coding control signal 205 instructing the coding of the current input image 201 and the recording retention of the extraction parameter 203 is output. Output. The threshold value may be set to a fixed value prior to a series of code decoding processes, or may be changed according to the severity of fluctuations in the extraction parameters.

【0030】

The diffifier 25 obtains the difference between the extraction parameter 203 and the selected reference parameter, and transmits the difference as the synthesis parameter 207 to the decoding side. Specifically, as shown in the example of (a) in Fig. 3, the extraction parameter is the center coordinate position g of the subject area._iAnd the inclination of the spindle θ_iThen the optimal reference parameter g_z, θ_zDifference from (d (g)_i), D (θ_i)) = (G_i-g_z, θ_i-θ_z) Let be the difference parameter 207. In addition, as shown in the example of (b) in Fig. 3, the extraction parameters are multiple feature points (p) in the subject area._i, q_i, r_i, s_i), Then the optimal reference parameter (p)_z, q_z, r_z, s_z) And the difference (d (p)_i), d (q_i), d (r)_i), d (s)_i)) = (p_i-p_z, q_i-q_z, r_i-r_z, s_i-s_z) Let be the difference parameter 207. Alternatively, as in the example of (c) in Fig. 3, the extraction parameter is a value (f) that represents the state of the local part of the subject area._i, er_i, rl_i, m_i), Then the optimal reference parameter (f)_z, er_z, el_z, m_z) And the difference (d (f)_i), d (er_i), d (el_i), d (m_i)) = (f_i-f_z, er_i-er_z, el_i-el_z, m_i-m_z) Let be the difference parameter 207. If the comparison circuit 23 does not output the reference control signal 206, the synthesis parameter 207 is not transmitted. The image coding circuit 21 encodes the current input image 201 according to the instruction of the coding control signal 205 and transmits the coded data 202 to the decoding side.

【0031】

The parameter memory 24 also additionally records the current extraction parameter 203 according to the instruction of the coding control signal 205. The parameter memory 24 records and holds the parameters extracted for each of the plurality of input images that have been code-decoded in the past, and outputs these as a plurality of reference parameters 204. If the parameter memory 204 is full and it is difficult to add the parameter memory 204 during the additional recording, an appropriate parameter that has already been recorded is deleted, and the extraction parameter 203 is additionally recorded. The parameter to be deleted is selected by using the history selected as the optimum reference parameter in the comparison circuit 23 by the method described above. Specifically, the comparison circuit 23 records the selected time for each reference parameter, and deletes the parameter having the oldest last selected time. Alternatively, the number of times selected may be accumulated for each reference parameter in a plurality of frame periods up to the present, and the parameter having the least number of selections may be deleted.

【0032】

The decoding side records and holds the coded data 202 received in the coded data memory 26. Further, the received coded data 202 is directly supplied to the image coding circuit 27. On the other hand, when the reference selection signal 206 is received instead of the coded data 202, the coded data memory 26 indicates one of the plurality of coded data recorded and held by the reference selection signal 206. Select and output as reference data 208. When holding the record, if the coded data memory 26 is already full and additional recording is difficult, delete one of the plurality of coded data already recorded and renew it. The coded data 202 is additionally recorded. As the method for selecting the coded data to be deleted, the same method as the method for selecting the parameter to be deleted in the parameter memory 24 described above is used. Further, the maximum number of coded data that can be recorded and held is also the same as the maximum number of parameters that can be recorded in the parameter memory 24 described above. With the above settings, the update of the record in the coded data memory 26 and the update of the record in the parameter memory 24 on the coding side can be automatically realized in synchronization with each other.

【0033】

The image decoding circuit 27 decodes any one of the supplied coded data 202 or reference data 208 and outputs the decoded image 209.

【0034】

The image composition circuit 28 performs deformation composition processing on the decoded image 209 using the received composition parameter 207, and outputs it as an output image 210 for display reproduction. Specifically, as described above using the example of FIG. 3A, the composite parameter 207 is the displacement amount d (g) of the central coordinate position of the subject area and the inclination of the main axis._i), D (θ_i), Then d (g) for each pixel that constitutes the subject area_i) Translation and front center position g_iAround d (θ_i) Rotates. Further, as described above using the example of (b) in FIG. 3, the composite parameter 207 is the displacement amount d (p) of the coordinate positions of the plurality of feature points in the subject area._i), d (g_i), d (r)_i), d (s)_i), The parallel movement represented by the following equation is performed for each pixel constituting the subject area.

【0035】

(lp × d (p)_i) + lq × d (q_i) + lr × d (r)_i) + Ls × d (s)_i)) / (Lp + lq + lr + ls) Here, (lp, lq, lr, ls) are the pixel of interest and the feature point (p)._i, q_i, r_i, s_i) Is a coefficient that is inversely proportional to the distance to each. Alternatively, as described above using the example of (c) in FIG. 3, the displacement amount d (f) in the state where the composite parameter 207 is the local portion of the subject region._i), d (er_i), d (el_i), d (m_i), The movement and transformation processing is executed for each area corresponding to the parameter element.

【0036】

The series of code decoding processes described above can also be realized by the configuration example shown in FIG. The moving image code decoding device of FIG. 5 has a feature amount extraction circuit 52 that extracts the feature amount of the input image as a parameter, and records and holds the extracted parameters according to the instruction of the coding control signal, and has a plurality of past. The parameter memory 54 that outputs the extracted parameter as a reference parameter in the input image of the above, and the reference parameter having the highest similarity by comparing the extracted parameter with the plurality of reference parameters are selected, and the value of the maximum similarity is obtained. The comparison circuit 53, which either outputs the coding control signal or transmits the reference control signal to the decoding side according to the above, and the input image are encoded and encoded according to the instruction of the coding control signal. An image coding circuit 51 for transmitting data to the decoding side, and a coding device consisting of a diffifier 55 for obtaining the difference between the extracted parameter and the selected reference parameter and transmitting the difference as a synthesis parameter to the decoding side. An image decoding circuit 56 that decodes the received series of the encoded data to generate a decoded image and outputs the decoded image as an output image for display / reproduction, and records and holds the decoded image and records the encoded data. When not received, the reference image is deformed and synthesized using the image memory 57 that selects one from the plurality of decoded images held and outputs it as a reference image according to the instruction of the reference control signal and the received synthesis parameters. It is composed of an image composition circuit 58 that generates an output image and outputs it for display reproduction.

【0037】

In the example of FIG. 5, on the coding side, the coded data 502 of the input image, the reference control signal 506, and the composite parameter 507 are output in the same manner as described above with reference to FIG. On the decoding side, the image decoding circuit 56 decodes the coded data 502 and outputs the decoded image 508. At the same time, the decoded image 508 is output as an output image 510 for display reproduction.

【0038】

The image memory 57 records and holds the decoded image 508. When the reference control signal 506 is received instead of the coded data 502, the image memory 57 selects and refers to one of the plurality of decoded images recorded and held by the reference selection signal 506. Output as image 509. When holding the record, if the image memory 57 is already full and additional recording is difficult, one of the plurality of decoded images already recorded is deleted, and the decoded image is newly recorded. Record additional 508. The method of selecting the decoded image to be deleted uses the same method as the method of selecting the parameter to be deleted in the parameter memory 54. Further, the maximum number of decoded images that can be recorded and retained is also the same as the maximum number of parameters that can be recorded in the parameter memory 54 described above. With the above settings, the update of the record of the image memory 57 and the update of the record of the parameter memory 54 on the coding side can be automatically realized in synchronization with each other.

【0039】

The image composition circuit 58 transforms and synthesizes the reference image 509 using the received composition parameter 507 and outputs it as an output image 510 for display reproduction.

【0040】

Further, in the example of FIG. 5, a plurality of reference parameters similar to the current extraction parameter 503 may be selected by the comparison circuit 53 on the coding side, and a plurality of reference control signals 506 and a plurality of synthesis parameters 507 may be transmitted. In this case, the decoding side reads out a plurality of reference images 509 instructed by the reference control signal 506 from the image memory 57, synthesizes one image 510 by using the synthesis parameter 507 in the image synthesis circuit 58, and outputs the image 510. An example of the operation when the plurality of reference parameters are selected will be described with reference to FIG.

【0041】

In the example of FIG. 6, parameters extracted from a plurality of past input images are stored in the parameter memory in advance as reference parameters. Current input image I (t_cParameter P (t) extracted in)_c) = (F_c, er_c, el_c, m_c) Is compared with the above reference parameter, and P (t)_c) Similar parameter P (t)₁), P (t₂), P (t₃) Is selected. This selection result is transmitted to the decoding side as the reference control signal 506 in FIG. Also, for each element of the current extraction parameters, select the one that most closely resembles the multiple reference parameters selected above and Sel (t)._c). Specifically, the current input image I (t)_c), The image I (t) at the center coordinate position of the face part₁) Parameter f1 is selected and the eye condition is image I (t)₃) Parameter (er₃, el₃) Is selected and the mouth condition is image I (t)₂) Parameter m₂Is selected. Furthermore, the above f_c, f₁Is a simple value that represents the coordinate position, so find the difference and Dif (f)_c). Sel (t_c) And Dif (f)_c) And the composite parameter 507 in FIG. 5 are transmitted to the decoding side. On the decoding side, in FIG. 5, the reference control signal 506 is used to display three reference images D (t) from the image memory 57.₁), D (t₂), D (t₃) Is read. Here D (t₁), D (t₂), D (t₃) Are the input images I (t)₁), I (t₂), I (t₃) Is a decoded image obtained by code decoding. In the image composition circuit 58, the three reference images D (t) are used by using the composition parameter 507.₁), D (t₂), D (t₃) Is combined and output as an output image 510. Specifically, refer to the entire face part Image D (t₁) And reference image D (t) on top of this₃) And the reference image D (t)₂) Is pasted together to synthesize the output image 510.

【0042】

[Effect of the invention]

As described above, in the present invention, the feature amount of the image encoded in the past is held as history information on the coding side, and a series of received coded data is recorded and held on the decoding side. If there is a history of code-decoding similar images to the input image in the past, only the control signal for selecting the similar image is transmitted, and the decoding side keeps the retained similar image as it is. Alternatively, it is transformed and combined and displayed. Therefore, when there is little change between frames on the input image, or when similar input images reappear after a certain frame period, the image quality is excellent and similar to the input image with only a small amount of data transmission. It is possible to display the natural image on the decoding side.

[Simple explanation of drawings]

[Figure 1]

It is a basic block diagram of the code decoding method of the 1st moving image of this invention.

[Figure 2]

It is a basic block diagram of the code decoding method of the 2nd moving image of this invention.

[Fig. 3]

It is a figure explaining the example of the feature amount of an image.

[Fig. 4]

It is a block diagram of one Example which realizes the code decoding method of the 1st moving image of this invention.

[Fig. 5]

It is a block diagram of one Example which realizes the code decoding method of the 2nd moving image of this invention.

[Fig. 6]

It is a figure explaining one Example of the code decoding method of the 2nd moving image of this invention.

[Explanation of symbols]

11,21,41,51 Image coding circuit 12,22,42,52 Feature extraction circuit 13,23,43,53 Comparison circuit 14,24,44,54 Parameter memory 15,26 Coded data memory 16,27,45,56 Image decoding circuit 25,55 diff 28,58 Image synthesis circuit 46,57 image memory 101,201,401,501 Input image 102,202,402,502 Coded data 103,203,403,503 Extraction parameters 104,204,404,504 Reference parameters 105,205,405,505 Coding control signal 106,206,406,506 Reference control signal 107,208 Reference data 108,210,408,510 Output image 207,507 Synthetic parameters 209,407,508 Decrypted image 509 Reference image