JP3974138B2 - Moving picture decoding apparatus and moving picture decoding method - Google Patents

Description

  The present invention relates to a moving image decoding apparatus and a moving image decoding method, and in particular, a moving image in which a digitized moving image input signal is subjected to resolution conversion according to a resolution determined by a resolution determination unit and is predictively encoded. A moving picture decoding apparatus that receives a signal and restores a moving picture, and a digitized moving picture input signal undergoes high-resolution or low-resolution conversion, receives a predictive-coded moving picture signal, The present invention relates to a moving picture decoding method for performing image restoration.

  As a moving picture coding technique, H.264 standardized by ITU (International Telecommunication Union) -T (Telecommunication Standardization Sector). There are moving picture hybrid coding schemes such as H.261 and MPEG (Moving Picture Experts Group) -1/2. In these encoding methods, one frame of an input image is divided into a plurality of blocks, motion compensation and orthogonal transformation are performed for each block, and entropy encoding is performed.

  In such a moving picture hybrid coding system, the amount of code information generated as a result of coding one frame when there is a scene change or when there is a portion of intense movement in the moving picture is one frame. It may occur that the transmission information amount of the assigned standard is exceeded. In such a case, the amount of generated code information is forcibly reduced to the reference transmission information amount. As a result, the image restored at the transmission destination has an extremely deteriorated image quality or extreme Frame dropping occurs and the image becomes very difficult to see.

  In order to avoid such problems, conventionally, when there is a scene change or when there is a part of a moving image that is intensely moved, the resolution of the input image is reduced to reduce the amount of generated code information. For example, Japanese Patent Application Laid-Open No. 7-95566 proposes a moving image encoding apparatus. In this apparatus, a prediction parameter (motion vector) is calculated for each block constituting the image, and the average size of the motion vectors of the entire image is obtained from their sizes. Then, based on this average value, the resolution to be encoded is determined, and the resolution of both the input image and the reference image is converted to the determined resolution before encoding.

  Japanese Patent Application Laid-Open No. 63-155896 also discloses an apparatus for encoding after converting the resolution of both an input image and a reference image to a low resolution.

  However, in the above-described conventional apparatus, encoding is performed after converting the resolution of both the input image and the reference image to the determined resolution, so that the encoding is performed at a low resolution from a state where the encoding is performed at a high resolution. When a transition is made to the state to be performed, all portions of the reference image from which the predicted image is created are converted to low resolution. For this reason, there is no temporal change such as a background image in a video conference, and even an image portion that is not related to an increase in the amount of transmission information is converted to a low resolution. As a result, these background portions are unnecessary. There was a problem that it was blurred and caused a great deterioration visually.

  The present invention has been made in view of the above points, and is a moving picture decoding apparatus and a moving picture that prevent a picture quality of a still region such as a background from being deteriorated when switching from a high resolution to a low resolution. An object of the present invention is to provide an image decoding method.

  In order to achieve the above object, the present invention provides a moving picture decoding apparatus as shown in FIG. The moving picture decoding apparatus includes a high-resolution image storage unit 22 that stores decoded decoded images at a high resolution, a decoded image storage unit 23 that stores decoded decoded images, and a reproduced prediction error. Generate a decoded image based on the signal, store the decoded image in the decoded image storage unit 23, and generate a decoded image that stores the decoded image in the high-resolution image storage unit 22 only when the resolution of the decoded image is high. When the resolution of the decoded image is low, the high resolution for converting the decoded image to a high resolution and recording it in the high resolution image storage means 22 only for the block that has been encoded. The image update means 29 is comprised.

  Also, the high resolution image update means 29 has a low resolution of the decoded image that is decoded only for a block that has been encoded and the motion vector is not zero or that has been subjected to intraframe encoding. When it is the resolution, the resolution of the decoded image is converted to a high resolution and recorded in the high resolution image accumulating means 22, and only for the block that is encoded and the motion vector is zero, When the resolution of the decoded image is low, the resolution of the prediction error signal reproduced by the inverse quantization / orthogonal inverse transform unit 27 is converted to a high resolution and stored in the high resolution image storage unit 22. The image stored in the high-resolution image storage means 22 is updated by adding to the image at the corresponding position.

The moving picture decoding apparatus configured as described above can improve the image quality by adding a few circuits to the conventional moving picture decoding apparatus.
That is, the restoration means 21, decoded image storage means 23, resolution conversion means 24, predicted image generation means 26, inverse quantization / orthogonal inverse conversion means 27, decoded image generation means 28, and image display means 30 in the figure have been conventionally performed. It is an existing circuit component. That is, the decoded image storage unit 23 stores the decoded image sent from the decoded image generation unit 28 regardless of the resolution level. The resolution conversion unit 24 converts the reference image stored in the decoded image storage unit 23 according to the resolution sent from the restoration unit 21 and sends the converted reference image to the predicted image generation unit 26. If the information indicating the high resolution is sent from the restoration means 21, the resolution conversion means 24 uses the reference image stored in the decoded image storage means 23 as it is, if the resolution is high resolution, the predicted image generation means 26. If the resolution is low, it is converted to a high resolution and sent to the predicted image generation means 26. On the other hand, if the information indicating the low resolution is sent from the restoration means 21, the resolution conversion means 24 generates the predicted image as it is if the reference image stored in the decoded image storage means 23 is low. If the resolution is high, it is converted to a low resolution and sent to the predicted image generation means 26. The predicted image generation unit 26 outputs a pixel value of zero as a predicted image and performs inter-frame coding for intra-frame coding based on the coding method restored by the restoration unit 21 for each predetermined block of the image. If so, based on the reference image, a predicted image is generated using the motion vector restored by the restoration unit 21 and output to the decoded image generation unit 28. The decoded image generating means 28 adds the predicted image to the prediction error signal reproduced by the inverse quantization / orthogonal inverse transform means 27 for each predetermined block to obtain a decoded image, and the decoded image is stored in the decoded image storage means 23. To accumulate. Note that the resolution of the reproduction prediction error signal and the predicted image here is the resolution restored by the restoration means 21.

Next, the operation based on the newly added component will be described.
First, even in the case of low resolution decoding, an attempt is made to enable high resolution images to be displayed on the screen for portions such as the background where there is no movement. On the other hand, decoded image storage means 23, resolution conversion means 24, The loop formed by the predicted image generation unit 26 and the decoded image generation unit 28 is a loop having the conventional configuration, and the operation in the loop is the same as the operation of the same loop on the transmission encoding side. There is a need. Therefore, when the resolution restored by the restoration unit 21 transitions from a low resolution to a high resolution, the resolution conversion unit 24 converts the low resolution image stored in the decoded image storage unit 23 to the high resolution. It is made to send to the prediction image generation means 26.

  Then, the predicted image generation unit 26 generates a predicted image, and the decoded image generation unit 28 generates a decoded image based on the predicted image, and stores the decoded image in the decoded image storage unit 23. At the same time, the decoded image is also stored in the high-resolution image storage means 22. The decoded image stored in the decoded image storage means 23 has a high resolution. If information indicating high resolution is continuously sent from the restoration means 21 after this transition, a high resolution image is continuously sent from the decoded image storage means 23 to the predicted image generation means 26.

  Next, when the resolution restored by the restoration unit 21 transitions from the high resolution to the low resolution, the resolution conversion unit 24 converts the high resolution image from the decoded image storage unit 23 to the low resolution, and the predicted image generation unit 26. As a result, the decoded image storage means 23 stores a low-resolution restored image. If information indicating low resolution is continuously sent from the restoration means 21 after this transition, a low resolution image is continuously sent from the decoded image storage means 23 to the predicted image generation means 26.

  When the resolution restored by the restoration unit 21 is a low resolution, the high resolution image update unit 29 decodes the low resolution obtained from the decoded image generation unit 28 only for the block that has been encoded. The image is converted to high resolution, and the image of the corresponding block stored in the high resolution image storage means 22 is updated. As a result of this update, only the image of the coding block having a change in the image among the images stored in the high-resolution image storage means 22 is updated, and the blocks constituting the background image without the change in the image are It is maintained as it is and high image quality is maintained. Accordingly, when the image stored in the high-resolution image storage means 22 is displayed on the image display means 30, the image quality of the background image is excellent in the image, and the background image that has conventionally occurred when switching the resolution is displayed. No change in image quality occurs.

According to the present invention, it is possible to prevent the image quality of a still region such as a background from being deteriorated when switching from a high resolution to a low resolution.
As a result, when switching from the low resolution to the high resolution, it is necessary in the prior art to newly transmit detailed information related to the image of the still area such as the background, but this is not necessary in the present invention.

  In addition, the coding block is divided into a moving area coding block and a still area coding block, and a high resolution is maintained for the still area coding block. As a result, it is possible to appropriately cope with a block which has been encoded by changing the luminance but has no motion in the image.

  Further, by simply adding a high-resolution image accumulating unit and a high-resolution image updating unit to the conventional video decoding device, the configuration of the conventional video decoding device is utilized and the resolution is changed from a high resolution to a low resolution. When switching, it is possible to prevent the image quality of a still region such as the background from being deteriorated.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
First, the principle configuration of the first embodiment will be described with reference to FIG. The first embodiment relates to a moving picture coding apparatus. The configuration includes a high-resolution image accumulating unit 3 that accumulates a locally decoded decoded image at a high resolution, a low-resolution image accumulating unit 4 that accumulates a locally decoded decoded image at a low resolution, and a resolution determining unit 1. When the high resolution is determined, the image stored in the high resolution image storage means 3 is read when the high resolution is determined, and the image stored in the low resolution image storage means 4 is determined when the low resolution is determined. A low-resolution high-resolution image that is recorded in the high-resolution image storage unit 3 by converting the image that has been predictively encoded at a low resolution and stored in the low-resolution image storage unit 4 into a high resolution. When the resolution determined by the update unit 13 and the resolution determination unit 1 changes from high resolution to low resolution, the image stored in the high resolution image storage unit 3 is converted to low resolution. , Consisting of a high-resolution at a low resolution image updating means 14 for recording the low resolution image storage unit 4.

  Further, the low resolution high resolution image update means 13 performs the low resolution only by using the resolution determination means 1 for blocks that have been encoded and whose motion vectors are not zero or that have undergone intraframe encoding. Is determined, the resolution of the image stored in the low-resolution image storage means 4 is converted to a high resolution, recorded in the high-resolution image storage means 3, and encoded, and Only when a low resolution is determined by the resolution determining means 1 for a block having a motion vector of zero, the resolution of the prediction error signal reproduced by the inverse quantization / orthogonal inverse transform means 11 is converted to a high resolution. Then, the image stored in the high resolution image storage unit 3 is updated by adding to the image at the corresponding position stored in the high resolution image storage unit 3.

  In the figure, resolution determination means 1, input image resolution conversion means 2, prediction parameter calculation means 6, prediction image generation means 7, prediction error signal generation means 8, orthogonal transform / quantization means 9, code allocation means 10, inverse quantization The orthogonal inverse transform unit 11 and the decoded image generation unit 12 are conventional circuit components.

  That is, the resolution determining unit 1 determines whether to perform encoding at high resolution or low resolution. The input image resolution conversion unit 2 converts the resolution of the input image to the resolution determined by the resolution determination unit 1. The prediction parameter calculation means 6 compares the image output from the input image resolution conversion means 2 with the image read by the selective reading means 5 for each predetermined block constituting the image, Whether to perform encoding or interframe encoding is determined, and a motion vector is calculated. The predicted image generation means 7 outputs a pixel value of zero as a predicted image based on the calculation result of the prediction parameter calculation means 6 for each predetermined block, if it is intra-frame coding, and if it is inter-frame coding, Based on the image read by the selective reading means 5, a prediction image is generated and output using the motion vector calculated by the prediction parameter calculation means 6.

  The prediction error signal generation unit 8 calculates a difference between the image output from the input image resolution conversion unit 2 and the prediction image for each predetermined block to generate a prediction error signal. The orthogonal transform / quantization unit 9 performs orthogonal transform and quantization on the prediction error signal generated by the prediction error signal generation unit 8. The code allocation means 10 is based on at least the output of the orthogonal transform / quantization means 9, the resolution determined by the resolution determination means 1, and the encoding method determined by the prediction parameter calculation means 6 and the calculated motion vector. The corresponding codes are extracted from the codes assigned in advance to the plurality of combinations and output. The inverse quantization / orthogonal inverse transform unit 11 performs inverse quantization and orthogonal inverse transform on the output of the orthogonal transform / quantization unit 9 to reproduce a prediction error signal. The decoded image generation means 12 adds a prediction image to the prediction error signal reproduced by the inverse quantization / orthogonal inverse transformation means 11 for each predetermined block to obtain a decoded image. When the high resolution is determined, the high resolution image storage unit 3 stores the high resolution, and when the low resolution is determined, the low resolution image storage unit 4 stores the high resolution.

  In the above configuration, it is assumed that the resolution determining unit 1 first determines that encoding is to be performed at a high resolution. In this case, the input image resolution conversion means 2 converts the resolution of the input image to a high resolution. Each switch shown in FIG. 1 is located on the side indicated by a broken line.

  Since the high resolution is determined, the selective reading means 5 reads the image stored in the high resolution image storage means 3 and sends it to the prediction parameter calculation means 6 and the prediction image generation means 7. The prediction parameter calculation unit 6 compares the high resolution input image output from the input image resolution conversion unit 2 with the high resolution reference image read from the high resolution image storage unit 3 by the selective reading unit 5; It is determined whether to perform intra-frame coding or inter-frame coding, and a motion vector is calculated. This process is performed for each predetermined block constituting the image. Based on the calculation result of the prediction parameter calculation unit 6, the predicted image generation unit 7 outputs a pixel value of zero as a predicted image if intra-frame coding, and selectively reads out if it is inter-frame coding. 5, based on the high-resolution reference image read from the high-resolution image storage unit 3, a prediction image is generated and output using the motion vector calculated by the prediction parameter calculation unit 6. This process is also performed for each predetermined block.

  The prediction error signal generation unit 8 calculates the difference between the high resolution input image output from the input image resolution conversion unit 2 and the prediction image generated by the prediction image generation unit 7 for each predetermined block. An error signal is generated and sent to the orthogonal transform / quantization means 9. The operations of the orthogonal transform / quantization means 9, the subsequent code assignment means 10, and the inverse quantization / orthogonal inverse transform means 11 are the same as those of the prior art.

  The decoded image generation unit 12 adds the prediction image generated by the prediction image generation unit 7 to the prediction error signal reproduced by the inverse quantization / orthogonal inverse transformation unit 11 for each predetermined block to obtain a decoded image. The decoded image is stored in the high-resolution image storage means 3.

  Next, it is assumed that the resolution determination unit 1 determines to perform encoding at a low resolution. In this case, the input image resolution conversion means 2 converts the resolution of the input image to a low resolution. Further, each switch shown in FIG. 1 is switched to the side indicated by the solid line.

  The high-resolution low-resolution image update means 14 only displays an image for one frame stored in the high-resolution image storage means 3 when the resolution determined by the resolution determination means 1 changes from high resolution to low resolution. The resolution is converted into a low resolution and stored in the low resolution image storage means 4. The selective reading means 5 reads the image stored in the low resolution image storage means 4 and sends it to the prediction parameter calculation means 6 and the prediction image generation means 7. As described above, the latest image is stored in the low-resolution image storage unit 4 when the resolution is changed from the high resolution to the low resolution. Using this, the prediction parameter calculation means 6 compares the low resolution reference image sent from the selective reading means 5 with the low resolution input image output from the input image resolution conversion means 2, The same processing as in the case of resolution is performed. The predicted image generation means 7, the prediction error signal generation means 8, the orthogonal transform / quantization means 9, the code allocation means 10, and the inverse quantization / orthogonal inverse transform means 11 also perform the same processing as in the case of high resolution.

  The decoded image generation unit 12 adds the prediction image generated by the prediction image generation unit 7 to the prediction error signal reproduced by the inverse quantization / orthogonal inverse transformation unit 11 for each predetermined block to obtain a decoded image. The decoded image is stored in the low-resolution image storage means 4. At this time, the high resolution image update means 13 at the time of low resolution converts the resolution of the image stored in the low resolution image storage means 4 to a high resolution only for the block that has been encoded, The image stored in the high resolution image storage means 3 is updated.

  That is, in moving picture coding, processing is performed in units of blocks constituting an image. As a result, however, it is divided into two types of blocks, coded blocks and non-coded blocks, due to differences in processing methods. The coding block is a block in which there is a change in the image when the previous image and the current image are compared, and the difference is naturally coded. Then, at the time of decoding the encoded block, image update is performed on the corresponding block of the previous reference image. On the other hand, the non-encoded block is a block in which there is no change between the previous image and the current image, and in this case, encoding is not performed. Therefore, when decoding a non-encoded block, the corresponding block of the previous reference image is not changed and the block is maintained as it is.

  Therefore, when the encoded block is processed, the high resolution image update unit 13 at the time of low resolution converts the resolution of the corresponding block image stored in the low resolution image storage unit 4 to a high resolution, The corresponding block image stored in the resolution image storage means 3 is updated. Such an update is not performed if an uncoded block is being processed. As a result, among the images stored in the high-resolution image storage means 3, only the image of the coding block having a change in the image is updated, and the blocks constituting the background image without the change in the image remain as they are. Maintained. As a result, the latest reference image is held in the high-resolution image storage unit 3 even though the low-resolution encoding process is performed. Since the encoding here is performed at a low resolution, the image quality of the updated block of the images stored in the high-resolution image storage means 3 is degraded. However, in a block that is not updated, such as a background image, high image quality is maintained. Therefore, after this, when the image is switched from the low resolution to the high resolution and the image stored in the high resolution image storage means 3 is used as a reference image, it is possible to prevent the deterioration of the background image quality as compared with the conventional technique. At the same time, transmission of detailed information related to the background, which was conventionally necessary at the time of switching, becomes unnecessary, and an increase in the amount of transmission information can be suppressed.

  More specifically, the high-resolution image update means 13 at the time of low resolution determines the resolution only for a block that has been encoded and the motion vector is not zero or has been subjected to intra-frame encoding. When the low resolution is determined by the means 1, the resolution of the image stored in the low resolution image storage means 4 is converted to a high resolution, and the image stored in the high resolution image storage means 3 is updated. . In addition, only the block that has been encoded and the motion vector is zero is reproduced by the inverse quantization / orthogonal inverse transform unit 11 when the resolution determination unit 1 determines the low resolution. The resolution of the prediction error signal is converted to a high resolution, and the converted reproduction prediction error signal is added to the image stored in the high resolution image storage means 3 to obtain a decoded image. The decoded image is used as the high resolution image. The image stored in the storage unit 3 is updated.

  That is, a block constituting an image such as a background without motion in the image should be processed as an uncoded block. However, in such a background image block, the brightness changes due to the movement of the shadow of the person, the change of the aperture of the camera, the flickering of the fluorescent lamp used for indoor lighting, noise, etc., resulting in a prediction error. Thus, encoding may be performed. In order to avoid such a situation, the coding block is divided into a stationary area coding block and a moving area coding block. The still area coding block is a block in the above case, and the moving area coding block is an original coding block.

The motion region coding block is selected based on the condition that the coding is performed and the motion vector is not zero or the intra-frame coding is performed.
The still area coding block is selected based on the condition that the coding is performed and the motion vector is zero. Only when the low resolution is determined by the resolution determining unit 1 only for the still area coding block, the resolution of the prediction error signal reproduced by the inverse quantization / orthogonal inverse transform unit 11 is converted to a high resolution, The converted reproduction prediction error signal is added to the image of the corresponding block stored in the high-resolution image storage unit 3 with high image quality to obtain a decoded image, and the decoded image is sent to the high-resolution image storage unit 3. Update the stored image. Thereby, even when the brightness of the background area changes, the image quality of the background area does not deteriorate.

Next, the first embodiment will be specifically described.
FIG. 3 is a block diagram showing a specific configuration of the first embodiment. Here, the resolution determining unit 1 shown in FIG. 1 corresponds to the resolution determining unit 31 in FIG. 3, and similarly, the input image resolution converting unit 2 is in the downsampling unit 32a and the changeover switch 32b, and the high resolution image accumulating unit 3 is in. The CIF (Common Intermediate Format) image storage unit 33, the low-resolution image storage unit 4 moves to the QCIF (Quarter Common Intermediate Format) image storage unit 34, the selective reading unit 5 moves to the changeover switch 35, and the prediction parameter calculation unit 6 moves. In the prediction unit 36, the predicted image generation means 7 is in the predicted image generator 37 a and the changeover switch 37 b, the prediction error signal generation means 8 is in the subtractor 38, and the orthogonal transform / quantization means 9 is in a DCT (Discrete Cosine Transform) 39 a and In the quantizer 39b, the code allocation means 10 is in the entropy encoder 40, the inverse quantization / orthogonal inverse transformation means 11 is in the inverse quantizer 41a and the IDCT (Inverse D iscrete Cosine Transform) 41b, the decoded image generating means 12 is added to the adder 42 and the changeover switch 45, and the high resolution image update means 13 is applied to the high resolution image update unit 43 and the changeover switch 45 at the time of low resolution. The low-resolution image update unit 14 corresponds to the downsampling unit 44 and the changeover switch 45.

In FIG. 3, a control unit 46 and an encoded information buffer 47 are further provided.
The resolution determination unit 31 receives the quantization step width from the control unit 46, the generated code information amount from the entropy encoder 40, and the occupancy information amount from the encoding information buffer 47. Based on these, excessive frame dropping is performed. In addition, an optimal resolution is determined so that a moving image can be transmitted in a state where image quality degradation does not occur. In this example, the input image is a high-resolution CIF (352 × 288 pixels per frame), and the resolution determination unit 31 normally determines that encoding is performed with the high-resolution CIF, and the transmission code information amount is a predetermined standard. Only when the amount is exceeded, it is determined that encoding is performed at a low resolution QCIF (176 × 144 pixels per frame). As the resolution determination method, for example, the method proposed in Japanese Patent Application No. 8-75605 can be applied, and Japanese Patent Application No. 8-75605 has a detailed description thereof. The determined resolution is sent to the changeover switches 32b, 35, 45 and the entropy encoder 40.

  The downsampling unit 32a performs 2-to-1 downsampling on the high resolution CIF image to create a low resolution QCIF image. Details of the downsampling process are shown in FIG.

  FIG. 4 is a diagram for explaining a 2-to-1 downsampling process. That is, in FIG. 4, white circles represent pixels constituting a high-resolution CIF image, and lowercase letters in the alphabet represent pixel values of the pixels. A black circle represents a pixel constituting a low-resolution QCIF image, and an uppercase letter of the alphabet represents a pixel value of the pixel. In order to obtain the pixel values A, B, C,... Of the pixels constituting the low resolution QCIF image, the pixel values of the four high resolution CIF pixels surrounding each of the low resolution QCIF pixels are used, for example, The average value thereof is obtained as A = (a + b + e + f) / 4. Thus, the conversion from the high resolution CIF to the low resolution QCIF is performed.

  Returning to FIG. 3, when the resolution determination unit 31 determines the high resolution CIF, the changeover switch 32 b is positioned on the side indicated by the broken line, and the input image of the high resolution CIF is selected, while the resolution determination unit 31. When the low resolution QCIF is determined, the changeover switch 32b is positioned on the side indicated by the solid line, and the low resolution QCIF image output from the downsampling unit 32a is selected.

  The changeover switch 35 operates according to the resolution sent from the resolution determination unit 31. When the high resolution CIF is being sent, the changeover switch 35 is located on the side indicated by the broken line, while when the low resolution QCIF is being sent, the changeover switch 35 is located on the side indicated by the solid line.

  The motion prediction unit 36 performs processing for each macro block (referred to as “block” in the present specification) constituting an image. A series of moving image encoding processing including this processing is H.264 which is a moving image encoding method standardized by ITU-T. Based on H.261.

  The predicted image generator 37 a generates a predicted image using the motion vector sent from the motion prediction unit 36 as the reference image sent from the changeover switch 35. Based on the encoding method sent from the motion prediction unit 36, the changeover switch 37b selects and outputs a pixel value 0 as a predicted image for each block if it is intra-frame encoding. For example, the prediction image generated by the prediction image generator 37a is selected and output.

  The DCT 39a performs discrete cosine transform on the prediction error signal to generate a transform coefficient. The quantizer 39b quantizes the transform coefficient according to the quantization step width sent from the control unit 46, and generates a quantized coefficient.

  The control unit 46 receives the amount of code information generated as a result of encoding from the entropy encoder 40 and the amount of occupation information from the encoding information buffer 47. Based on these, the control unit 46 determines the quantization step width and sends the quantization step width to the quantizer 39b, the inverse quantizer 41a, the resolution determination unit 31, and the entropy encoder 40.

  The entropy encoder 40 includes the quantization coefficient output from the quantizer 39b, the resolution determined by the resolution determination unit 31, the quantization step width determined by the control unit 46, the encoding method determined by the motion prediction unit 36, and Based on the motion vector, a corresponding code is extracted from codes assigned in advance to the plurality of combinations, and output to the encoded information buffer 47. The encoding information buffer 47 temporarily stores the code output from the entropy encoder 40.

  The downsampling unit 44 has the same configuration as the downsampling unit 32a. The changeover switch 45 operates according to the resolution sent from the resolution determination unit 31. When the high resolution CIF is being sent, each intercept of the changeover switch 45 is located on the side indicated by a broken line, while when the low resolution QCIF is being sent, each intercept of the changeover switch 45 is on the side indicated by a solid line. Each is located. However, the changeover switch 45 connects the output of the downsampling unit 44 to the QCIF image storage unit 34 only immediately after the transition from the high resolution CIF to the low resolution QCIF. Therefore, the latest QCIF image is stored in the QCIF image storage unit 34 immediately after the transition from the high resolution CIF to the low resolution QCIF.

Next, operations of the high-resolution image update unit 43 and the changeover switch 45 at the time of low resolution will be described with reference to FIGS.
FIG. 5 is a diagram illustrating an operation of the high-resolution image update unit 43 at the time of low resolution when a moving area coding block is a processing target. That is, when the decoding process is performed with the low resolution QCIF, as shown in FIG. 3, each intercept of the changeover switch 45 is located on the side indicated by the solid line. Then, the high-resolution image update unit 43 at the time of low resolution is sent information about the encoding method and the motion vector from the motion prediction unit 36, and based on these, the block to be processed is encoded, and Then, it is determined whether or not the motion vector is non-zero or the motion area coding block is subjected to intra-frame coding. When the moving area coding block is a processing target, the adder 42 adds the low-resolution reproduction prediction error signal sent from the IDCT 41b to the low-resolution prediction picture sent from the changeover switch 37b. Output a low resolution decoded image. The decoded image is sent to the QCIF image storage unit 34 via the changeover switch 45 (switch position indicated by a solid line in FIG. 3).

  FIG. 7A is a diagram illustrating an image for one frame stored in the QCIF image storage unit 34. In the figure, shaded blocks are encoded blocks, which are portions where there is motion in the image. The non-encoded block corresponds to an image having no motion such as a background. Since the image stored in the QCIF image storage unit 34 has a low resolution, the image quality of any block image is not good.

  The low resolution high resolution image update unit 43 includes an upsampling unit 43a. The upsampling unit 43a performs one-to-two upsampling on the low resolution QCIF image to create a high resolution CIF image. Details of the upsampling process are shown in FIG.

  FIG. 8 is a diagram for explaining a one-to-two upsampling process. That is, in FIG. 8, black circles represent pixels constituting a low-resolution QCIF image, and uppercase letters in the alphabet represent pixel values of the pixels. A white circle represents a pixel constituting a high-resolution CIF image, and a small letter of the alphabet represents a pixel value of the pixel. A broken line indicates a boundary between blocks. Here, the pixel values of the pixels constituting the high-resolution CIF image not in contact with the boundary line, for example, f, are used as the pixel values of the pixels constituting the four low-resolution QCIF images close to the high-resolution CIF pixel. Thus, f = (9A + 3B + 3C + D) / 16 is obtained by weighting according to the proximity. However, the pixel value of the pixel of the high resolution CIF that is in contact with the boundary line, for example, b, does not use the pixel value of the pixel of the adjacent block, but two low resolutions on the own block side close to the pixel of the high resolution CIF. Using the pixel value of the QCIF pixel, b = (3A + B) / 4 is obtained. Similarly, for example, i does not use the pixel value of the pixel in the adjacent block, but uses the pixel values of the two low-resolution QCIF pixels on the own block side close to the high-resolution CIF pixel. A + 3C) / 4. For a located at the corner of the boundary line, a = A is obtained.

  Returning to FIG. 5, when the high resolution image update unit 43 at the time of low resolution determines that the moving area coding block is the processing target, the processing among the images stored in the QCIF image storage unit 34 is performed. Upsampling is performed on the target block. That is, if the coding block shown in FIG. 7A is a moving area coding block, this block is taken out and converted to a high resolution CIF. Then, the image is sent to the CIF image storage unit 33 via the changeover switch 45 (switch position indicated by a solid line in FIG. 3), and the image of the corresponding block is updated.

  FIG. 7B is a diagram illustrating an image for one frame stored in the CIF image storage unit 33. In the figure, the shaded blocks are blocks that have been updated by the high-resolution image update unit 43 at the time of low resolution, and are portions that have motion in the image. The non-update block corresponds to an image having no motion such as a background. Since such an update is performed, the latest image is held in the CIF image storage unit 33 even when the low resolution QCIF processing is performed. Since the image of the updated block is updated using the low-resolution QCIF image, the image quality of the block is reduced, but the image stored in the non-update block can maintain the high image quality.

  FIG. 6 is a diagram illustrating an operation of the high-resolution image update unit 43 at the time of low resolution when a still area encoded block is a processing target. That is, when the decoding process is performed with the low resolution QCIF, as shown in FIG. 3, each intercept of the changeover switch 45 is located on the side indicated by the solid line. Then, the high-resolution image update unit 43 at the time of low resolution receives information on the encoding method and the motion vector from the motion prediction unit 36, and based on these, the block to be processed is encoded (quantum The number of quantized coefficients output from the encoder 39b is not zero), and it is determined whether or not it is a still area coding block having a motion vector of zero. When the still area coding block is the processing target, the adder 42 adds the low-resolution reproduction prediction error signal sent from the IDCT 41b to the low-resolution prediction image sent from the changeover switch 37b. Output a low resolution decoded image. The decoded image is sent to the QCIF image storage unit 34 via the changeover switch 45 (switch position indicated by a solid line in FIG. 3).

  The low resolution high resolution image update unit 43 includes an upsampling unit 43b and an adder 43c in addition to the upsampling unit 43a. The upsampling unit 43b performs one-to-two upsampling on the low-resolution QCIF reproduction prediction error signal to create a high-resolution CIF reproduction prediction error signal. The configuration of the upsampling unit 43b is the same as that of the upsampling unit 43a.

  When the low-resolution high-resolution image update unit 43 determines that the still area encoded block is a processing target, the up-sampling unit 43b up-samples the reproduction prediction error signal sent from the IDCT 41b. The adder 43c adds the upsampled reproduction prediction error signal to the high resolution CIF image of the corresponding block stored in the CIF image storage unit 33 to obtain a high resolution CIF decoded image. Is sent to the CIF image storage unit 33 via the changeover switch 45 (switch position indicated by a solid line in FIG. 3), and the image of the corresponding block in the CIF image storage unit 33 is updated. Since such an update is performed, the latest image is held in the CIF image storage unit 33 even when the low resolution QCIF processing is performed. At the same time, since the decoded image is created using the high-resolution CIF image, the image quality of the image of the updated block is not deteriorated, and the image stored in the updated block has high image quality. Can be maintained. Moreover, the image takes into account changes in luminance.

If an image display device is connected to the CIF image storage unit 33, an image with a high-quality background maintained can be monitored by this image display device even if the resolution changes.
As described above, in the first embodiment, even when the resolution is lowered so that the transmission information amount does not exceed the predetermined reference amount, the background image without movement in the image stored in the CIF image storage unit 33 is used. Can maintain high image quality. Therefore, when the image stored in the CIF image storage unit 33 is monitored, the background image is not blurred even when the resolution changes from the high resolution to the low resolution. This is effective in a video decoding apparatus that receives and decodes an encoded signal sent from such a video encoding apparatus, as shown in a second embodiment described later. Also, when the resolution changes from low resolution to high resolution, the amount of transmission information increases in order to restore the low-quality background image to high image quality, but the CIF image storage unit 33 has high image quality. Therefore, such an increase in the amount of transmission information can be prevented beforehand.

Next, a second embodiment will be described.
The second embodiment relates to a moving picture decoding apparatus.
FIG. 9 is a block diagram illustrating a configuration of the second embodiment. In the second embodiment, the configuration of the first embodiment is applied. Therefore, in FIG. 9, the same reference numerals are given to the same parts as those of the first embodiment, and the description thereof is omitted.

  In the second embodiment, the local decoding unit 50 has the same configuration as that of the first embodiment. An entropy decoder 51 and an image display device 52 are newly provided. The entropy decoder 51 restores the quantization coefficient, the resolution, the quantization step width, the encoding method, and the motion vector based on the encoded signal sent from the transmission side. Then, the quantization coefficient and the quantization step width are transferred to the inverse quantizer 41a, the resolution is transferred to the changeover switches 35 and 45, the encoding method and the motion vector are set to the prediction image generator 37a, the changeover switch 37b, and the high resolution at the low resolution. The image is sent to the image update unit 43.

  The encoded signal sent from the moving image encoding apparatus as shown in the first embodiment is received by the moving image decoding apparatus, and the local decoding unit 50 is the same as the same components in the first embodiment. Perform the same operation. As a result, the CIF image storage unit 33 always holds an image in which a high-quality background is maintained even when the resolution changes. This image is displayed. Therefore, even when the resolution is lowered so that the transmission information amount does not exceed the predetermined reference amount, the high-quality image can be maintained for the background image in which the image does not move, thereby changing the resolution from the high resolution to the low resolution. Even in the case, the background image is not blurred in the image display device 52 on the receiving side.

Next, a third embodiment will be described.
First, the principle configuration of the third embodiment will be described with reference to FIG. The third embodiment relates to a moving picture decoding apparatus. The configuration is based on the high-resolution image storage means 22 for storing the decoded decoded image at high resolution, the decoded image storage means 23 for storing the decoded decoded image, and the reproduced prediction error signal. A decoded image generating unit 28 for obtaining a decoded image, storing the decoded image in the decoded image storage unit 23, and storing the decoded image in the high-resolution image storage unit 22 only when the resolution of the decoded image is high; When the resolution of the decoded image is a low resolution, the high-resolution image update unit 29 converts the decoded image to a high resolution and records the decoded image in the high-resolution image storage unit 22 only for the block that has been encoded. It consists of.

  Also, the high resolution image update means 29 has a low resolution of the decoded image that is decoded only for a block that has been encoded and the motion vector is not zero or that has been subjected to intraframe encoding. When it is the resolution, the resolution of the decoded image is converted to a high resolution, recorded in the high resolution image storage means 22, and the decoding is performed only for the block that is encoded and the motion vector is zero. When the resolution of the decoded image is low, the resolution of the prediction error signal reproduced by the inverse quantization / orthogonal inverse transform unit 27 is converted to a high resolution and stored in the high resolution image storage unit 22. The image stored in the high-resolution image storage means 22 is updated by adding to the image at the corresponding position.

The moving picture decoding apparatus configured as described above only adds a few circuits to the conventional moving picture decoding apparatus.
That is, the restoration unit 21, the decoded image storage unit 23, the resolution conversion unit 24, the predicted image generation unit 26, the inverse quantization / orthogonal inverse conversion unit 27, the decoded image generation unit 28, and the image display unit 30 are existing circuits. It is a component part. That is, the decoded image storage unit 23 stores the decoded image sent from the decoded image generation unit 28 regardless of the resolution level. The resolution conversion unit 24 converts the reference image stored in the decoded image storage unit 23 according to the resolution sent from the restoration unit 21 and sends the converted reference image to the predicted image generation unit 26. If the information indicating the high resolution is sent from the restoration means 21, the resolution conversion means 24 uses the reference image stored in the decoded image storage means 23 as it is, if the resolution is high resolution, the predicted image generation means 26. If the resolution is low, it is converted to a high resolution and sent to the predicted image generation means 26. On the other hand, if the information indicating the low resolution is sent from the restoration means 21, the resolution conversion means 24 generates the predicted image as it is if the reference image stored in the decoded image storage means 23 is low. If the resolution is high, it is converted to a low resolution and sent to the predicted image generation means 26. The predicted image generation unit 26 outputs a pixel value of zero as a predicted image and performs inter-frame coding for intra-frame coding based on the coding method restored by the restoration unit 21 for each predetermined block of the image. If so, based on the reference image, a predicted image is generated using the motion vector restored by the restoration unit 21 and output to the decoded image generation unit 28. The decoded image generating means 28 adds the predicted image to the prediction error signal reproduced by the inverse quantization / orthogonal inverse transform means 27 for each predetermined block to obtain a decoded image, and the decoded image is stored in the decoded image storage means 23. To accumulate. Note that the resolution of the reproduction prediction error signal and the predicted image here is the resolution restored by the restoration means 21.

Next, the operation based on the newly added component will be described.
First, even in the case of low resolution decoding, an attempt is made to enable high resolution images to be displayed on the screen for portions such as the background where there is no movement. On the other hand, decoded image storage means 23, resolution conversion means 24, The loop formed by the predicted image generation unit 26 and the decoded image generation unit 28 is a loop having the conventional configuration, and the operation in the loop is the same as the operation of the same loop on the transmission encoding side. There is a need. Therefore, when the resolution restored by the restoration unit 21 transitions from a low resolution to a high resolution, the resolution conversion unit 24 converts the low resolution image stored in the decoded image storage unit 23 to the high resolution. It is made to send to the prediction image generation means 26.

  Then, the predicted image generation unit 26 generates a predicted image, and the decoded image generation unit 28 generates a decoded image based on the predicted image, and stores the decoded image in the decoded image storage unit 23. At the same time, the decoded image is also stored in the high-resolution image storage means 22. The decoded image stored in the decoded image storage means 23 has a high resolution. If information indicating high resolution is continuously sent from the restoration means 21 after this transition, a high resolution image is continuously sent from the decoded image storage means 23 to the predicted image generation means 26.

  Next, when the resolution restored by the restoration unit 21 transitions from the high resolution to the low resolution, the resolution conversion unit 24 converts the high resolution image from the decoded image storage unit 23 to the low resolution, and the predicted image generation unit 26. As a result, the decoded image storage means 23 stores a low-resolution restored image. If information indicating low resolution is continuously sent from the restoration means 21 after this transition, a low resolution image is continuously sent from the decoded image storage means 23 to the predicted image generation means 26.

  When the resolution restored by the restoration unit 21 is a low resolution, the high resolution image update unit 29 decodes the low resolution obtained from the decoded image generation unit 28 only for the block that has been encoded. The image is converted to high resolution, and the image of the corresponding block stored in the high resolution image storage means 22 is updated. As a result of this update, only the image of the coding block having a change in the image among the images stored in the high-resolution image storage means 22 is updated, and the blocks constituting the background image without the change in the image are It is maintained as it is and high image quality is maintained. Accordingly, when the image stored in the high-resolution image storage means 22 is displayed on the image display means 30, the image quality of the background image is excellent in the image, and the background image that has conventionally occurred when switching the resolution is displayed. No change in image quality occurs.

Next, the third embodiment will be specifically described.
FIG. 10 is a block diagram showing a specific configuration of the third embodiment. Here, the restoring means 21 shown in FIG. 2 corresponds to the code decoder 61 of FIG. 10, and similarly, the high resolution image storage means 22 is in the high resolution image storage section 62 and the decoded image storage means 23 is in the frame memory 63. The resolution conversion means 24 is in the CIF / QCIF conversion section 64, the predicted image generation means 26 is in the predicted image generation section 66, the inverse quantization / orthogonal inverse transformation means 27 is in the inverse quantizer 67a and the IDCT 67b, and the decoded image generation means. 28 corresponds to the adder 68 a and the switch 68 b, the high resolution image update unit 29 corresponds to the high resolution image update unit 69 and the switch 71, and the image display unit 30 corresponds to the image display device 70.

  The code decoder 61 restores the quantization coefficient, the resolution, the quantization step width, the encoding method, and the motion vector based on the encoded signal sent from the transmission side. Then, the quantization coefficient and the quantization step width are transferred to the inverse quantizer 67a, the resolution is transferred to the switches 68b and 71 and the CIF / QCIF converter 64, the encoding method and the motion vector are input to the predicted image generator 66 and the high resolution image. Send to the update unit 69.

  The CIF / QCIF converter 64 converts the resolution of the image read from the frame memory 63 into the resolution sent from the code decoder 61 and outputs it regardless of the resolution. If the high resolution CIF is sent, the resolution of the image read from the frame memory 63 is converted into the high resolution CIF, and if the low resolution QCIF is sent, it is converted into the low resolution QCIF and output.

  Only when the resolution sent from the code decoder 61 is the high resolution CIF, the switch 68b stores the decoded image of the high resolution CIF for each block sent from the adder 68a in the corresponding block of the high resolution image storage unit 62. To do. Therefore, the high-resolution image storage unit 62 holds the high-resolution CIF decoded image while the high-resolution CIF is being decoded. Hereinafter, a stored image of the high-resolution image storage unit 62 during a period when the decoding process is performed with the low-resolution QCIF will be described.

  FIG. 11 is a diagram illustrating an internal configuration of the high-resolution image update unit 69. That is, the high-resolution image update unit 69 includes up-sampling units 69a and 69b, an adder 69c, and a changeover switch 69d. Each of the upsampling units 69a and 69b has the same configuration as the upsampling unit of the first embodiment shown in FIG. The changeover switch 69d determines, based on the encoding method and motion vector sent from the encoder / decoder 61, whether the processing target block is a motion area encoding block or a still area encoding block, If it is a coding block, it is located on the side indicated by a broken line in FIG. 11, and if it is a still area coding block, it is located on the side indicated by a solid line in FIG. Note that the definitions of the moving area coding block and the still area coding block are the same as those shown in the first embodiment.

  In FIG. 10, the switch 71 operates in accordance with the resolution sent from the encoder / decoder 61, and the output of the high-resolution image update unit 69 is output only when the low-resolution QCIF is sent. Send to 62.

  Returning to FIG. 11, the decoding process when the low-resolution QCIF is being sent will be described. First, when the changeover switch 69d determines that the moving area coding block is the processing target, the intercept of the changeover switch 69d is located on the side indicated by the broken line in FIG. In the high-resolution image update unit 69, the up-sampling unit 69a performs up-sampling on the low-resolution QCIF decoded image sent from the adder 68a, creates a high-resolution CIF image, and passes through the changeover switch 69d. The image is sent to the high resolution image storage unit 62. Then, the image of the corresponding block is updated.

  Since such an update is performed, the latest image is held in the high-resolution image storage unit 62 even when the decoding process with the low-resolution QCIF is performed. Since the image of the updated block is updated using the low-resolution QCIF image, the image quality of the block is deteriorated, but the image stored in the non-update block can maintain the high image quality.

  Next, when the changeover switch 69d determines that the still area coding block is the processing target, the intercept of the changeover switch 69d is positioned on the side indicated by the solid line in FIG. In the high-resolution image update unit 69, the up-sampling unit 69b performs up-sampling on the low-resolution QCIF reproduction prediction error signal sent from the IDCT 67b, and creates a high-resolution CIF reproduction prediction error signal. The adder 69c adds the upsampled reproduction prediction error signal to the high-resolution CIF image of the corresponding block stored in the high-resolution image storage unit 62 to obtain a high-resolution CIF decoded image. The image is sent to the high-resolution image storage unit 62 via the changeover switch 69d and the switch 71, and the image of the corresponding block in the high-resolution image storage unit 62 is updated. By performing such an update, the latest image is held in the high-resolution image storage unit 62 even when the decoding process with the low-resolution QCIF is performed. At the same time, the image quality of the updated block image does not deteriorate because the decoded image is created using the high-resolution CIF image, and the image stored in the updated block has a high image quality. Can be maintained. Moreover, the image takes into account changes in luminance.

  In this way, an image whose background is maintained at a high image quality even when there is a change in resolution is always held in the high-resolution image storage unit 62. Therefore, as shown in FIG. 10, the image display device 70 displays the image stored in the high-resolution image storage unit 62. Therefore, even when the resolution is lowered so that the transmission information amount does not exceed the predetermined reference amount, the high-quality image can be maintained for the background image in which the image does not move, thereby changing the resolution from the high resolution to the low resolution. Even at times, the image display device 70 does not blur the background image.

  The moving picture decoding apparatus according to the third embodiment makes use of the configuration of the conventional moving picture decoding apparatus, and further includes a high resolution image storage unit 62, a high resolution image update unit 69, a switch 71, and a changeover switch. The image quality can be prevented from being lowered by adding only a small circuit 68b.

It is a 1st principle explanatory drawing. It is a 2nd principle explanatory drawing. It is a block diagram which shows the specific structure of 1st Embodiment. It is a figure explaining the downsampling process of 2 to 1. It is a figure which shows operation | movement of the high resolution image update part at the time of the low resolution in case a moving region encoding block is a process target. It is a figure which shows operation | movement of the high resolution image update part at the time of the low resolution in case a still area encoding block is a process target. (A) is a diagram illustrating an image for one frame stored in the QCIF image storage unit, and (B) is a diagram illustrating an image for one frame stored in the CIF image storage unit. is there. It is a figure explaining the upsampling process of 1 to 2. It is a block diagram which shows the structure of 2nd Embodiment. It is a block diagram which shows the specific structure of 3rd Embodiment. It is a figure which shows the internal structure of a high-resolution image update part.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Resolution determination means 2 Input image resolution conversion means 3 High resolution image storage means 4 Low resolution image storage means 5 Selective reading means 6 Prediction parameter calculation means 7 Prediction image generation means 8 Prediction error signal generation means 9 Orthogonal transformation / quantization means DESCRIPTION OF SYMBOLS 10 Code allocation means 11 Dequantization / orthogonal inverse transformation means 12 Decoded image generation means 13 Low resolution high resolution image update means 14 High resolution low resolution image update means 21 Restoration means 22 High resolution image storage means 23 Decoded image storage means 24 Resolution conversion means 26 Predicted image generation means 27 Inverse quantization / orthogonal inverse conversion means 28 Decoded image generation means 29 High resolution image update means 30 Image display means

Claims (4)

In a video decoding device that performs video decoding processing,
Restoring means for obtaining information indicating whether the received image information is low resolution or high resolution;
High-resolution image storage means for storing the decoded image at high resolution;
When the resolution acquired by the restoration means is high resolution , decoding means for storing a decoded image in the high resolution image storage means,
When the resolution acquired by the restoration means is low resolution, non-encoding of a still area where the image is unchanged and is not encoded with respect to the block constituting the high-resolution image corresponding to the decoded decoded image block High-resolution image update means for maintaining the block as a high-resolution image, converting only the encoded block that has been encoded into high-resolution, and recording and updating the high-resolution image storage means;
A moving picture decoding apparatus characterized in that an image stored in the high-resolution image storage means is output as a decoded image. The high-resolution image updating unit performs inverse quantization when the resolution obtained by the restoration unit is low for a still region coding block that has been encoded and has a motion vector of zero. / The resolution of the prediction error signal reproduced by the inverse orthogonal transform function is converted to a high resolution, and the converted reproduction prediction error signal is added to the image of the corresponding block stored in the high resolution image storage means. The moving picture decoding apparatus according to claim 1. In a video decoding method for performing video decoding processing,
Obtain information indicating whether the received image information is low resolution or high resolution,
The decoded image is stored at high resolution in the high resolution image storage means,
If the acquired resolution is high resolution , the decoded image is stored in the high resolution image storage means,
When the acquired resolution is low resolution, the non-encoded block in the still area where the image has not changed and is not encoded is compared with the block that constitutes the high resolution image corresponding to the decoded decoded image block. And only the encoded blocks that are encoded are converted to high resolution and recorded in the high resolution image storage means,
Outputting the image stored in the high-resolution image storage means as a decoded image;
A moving picture decoding method characterized by the above.
The high resolution image storage means, the encoding is performed, and, for the still region encoded block motion vector is zero, when the resolution is a low resolution, by the function of the inverse quantization / inverse orthogonal transformation The resolution of the reproduced prediction error signal is converted to a high resolution, and the converted reproduction prediction error signal is added to the image of the corresponding block of the high resolution image stored in the high resolution image storage means. The moving picture decoding method according to claim 3.

Images