JP4269405B2 - Encoding apparatus and method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an encoding apparatus and method suitable for use in an MPEG encoder that compresses and encodes an image by the MPEG method.
[0002]
[Prior art]
In digital satellite broadcasting, the MPEG2 system is used as an image compression system. In the MPEG2 system, motion compensation predictive coding using a motion vector is performed. That is, in the MPEG2 system, three types of screens called I (Intra) picture, P (Predicti) picture, and B (Bidirectionally Predictive) picture are sent. In an I picture, DCT encoding is performed using pixels on the same screen (frame or field). In a P picture, a motion battle is detected with reference to an already encoded I picture or P picture, and DCT encoding using motion compensated prediction is performed. In the B picture, DCT encoding using motion prediction is performed with reference to the preceding and subsequent I or P pictures.
[0003]
[Problems to be solved by the invention]
As described above, in the MPEG2 system, motion compensation predictive encoding is performed, and in an I picture or B picture, a motion vector is detected by referring to an already encoded I picture or P picture, and encoding is performed. Done. When such motion compensation predictive coding is performed, the amount of codes is reduced, and efficient compression processing can be performed.
[0004]
However, when performing motion compensation predictive coding, the reliability of the motion vector is important. If the reliability of the motion vector is low, the correlation between the reference picture and the current picture becomes weak, and if the code amount is lowered, the image quality is deteriorated.
[0005]
Possible cases where the reliability of the motion vector is low include a case where a motion vector is detected in a macroblock near the end of the screen, or a case where a motion vector is detected where the input image is difficult to encode. .
[0006]
That is, the motion vector extracts a block having the same size as that of the block of the current screen to be processed from the same position on the reference screen as a base point, and moves this block within a predetermined search range. Then, the absolute value sum of the difference between each pixel of the block on the current screen and each pixel of the block on the reference screen is obtained, and this is obtained by block matching such that the block having the minimum value is detected.
[0007]
In such block matching processing, since the search range is within the screen near the center of the screen, the reliability of the motion vector is high. However, in the area close to the end of the screen, the search range is outside the screen, so that the reliability of the motion vector is considered to decrease.
[0008]
Further, when the pattern is complex and the motion is intense, it is difficult to find a block in which the sum of absolute values of differences between each pixel of the current frame block and each pixel of the reference frame block is extremely small. For this reason, it is considered that the reliability of a motion vector detected in a place where the pattern is complicated, the motion is intense, and the encoding is difficult is reduced.
[0009]
As described above, when encoding is performed by performing motion compensation using a motion vector, if the reliability of the motion vector is low, image quality deteriorates at that portion.
[0010]
Accordingly, an object of the present invention is to provide an encoding apparatus and method which prevent deterioration of image quality in a portion where the reliability of a motion vector is low.
[0011]
[Means for Solving the Problems]
This invention Motion vector detecting means for obtaining a motion vector between the input image and the reference image;
Orthogonal transform means for performing motion compensation based on the motion vector obtained by the motion vector detection means to orthogonally transform the difference between the reference image and the input image;
Quantization means for quantizing the output of the orthogonal transform means;
If the difficulty of the input image exceeds a predetermined value, the quantized value control means for reducing the quantized value of the quantizing means, assuming that the reliability of the motion vector detected by the motion vector detecting means is low;
With
The difficulty of the input image is an encoding device that is an average quantization value at the time of the latest encoding.
[0012]
In this invention, the quantization value control means is a macroblock located near the end of the input image. For The quantization value of the quantization means is assumed to be unreliable. Make smaller .
[0013]
In the present invention, the quantized value control means divides the distance from the end of the input image into a plurality of pieces, and changes the quantized value of the quantizing means stepwise in accordance with the distance from the end of the input image.
[0014]
In this invention, the quantized value control means is a macroblock that performs intra-screen coding. For The quantization value is not changed regardless of the reliability of the motion vector.
[0017]
In this invention, the quantization value control means compares an index indicating the difficulty of encoding the input image with a plurality of threshold values, and changes the quantization value of the quantization means stepwise in accordance with the threshold values.
[0019]
In the macro block near the end of the screen, the search range is outside the screen, so that the reliability of the motion vector is considered to decrease. For this reason, in the macro block near the end of the screen, the quantization value is lowered and the code amount is increased in the portion where the reliability of the motion vector is considered to be lowered. This compensates for image quality degradation in a portion where the reliability of the motion vector near the end of the screen is reduced.
[0020]
Further, it is considered that the reliability of the motion vector is lowered where the difficulty of encoding the input image is great. Where the difficulty of encoding the input image is large, the pattern is complicated and the movement is large, and this can be judged from the latest quantized value. For this reason, in the portion where the latest average quantized value is equal to or greater than the predetermined value and the reliability of the motion vector is considered to be lowered, the quantized value is lowered and the code amount is increased. This compensates for image quality degradation in a portion where the reliability of the motion vector near the end of the screen is reduced.
[0021]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows an example of the configuration of an MPEG-2 encoder to which the present invention is applied. In an example of an MPEG2 system encoder to which the present invention is applied, since the macroblock at the end of the screen is considered to have low motion vector reliability, the macroblock at the end of the screen has a lower quantized value to reduce motion. Compensation for image quality deterioration due to a decrease in vector reliability is made.
[0022]
In FIG. 1, a component digital video signal including a luminance signal Y and color difference signals Cb and Cr is supplied to an input terminal 1. The digital video signal from the input terminal 1 is temporarily stored in the frame memory 2. The frame memory 2 has a capacity capable of storing at least three frames of images of the current screen, the past reference screen, and the future reference screen.
[0023]
Using the data stored in the frame memory 2, the motion vector detection circuit 3 obtains a motion vector between the reference screen and the current screen. The motion vector is obtained in units of macroblocks. In other words, the motion vector is extracted from a block of the current screen (frame or field) to be processed by extracting a block having the same size as that block from the same position on the reference screen as a base point. While moving this block, the absolute value sum of the difference between each pixel of the block on the current screen and each pixel of the block on the reference screen is obtained, and the block matching is performed such that the block having the minimum value is detected.
[0024]
The motion vector MV obtained by the motion vector detection circuit 3 is supplied to the variable length coding circuit 8 and the motion compensation circuit 15.
[0025]
Further, residual information e obtained when the motion vector detection circuit 3 obtains a motion vector is supplied to the mode setting circuit 4. Further, the position information of the macroblock when the motion vector detection circuit 3 obtains the motion vector is supplied to the terminal macroblock detection circuit 21. The end macroblock detection circuit 21 determines whether a macroblock at the end position on the screen is being obtained.
[0026]
In the I picture, the switch circuit 5 is switched according to the encoding mode by the mode setting signal from the mode setting circuit 4. That is, the switch circuit 5 is set on the terminal 5A side in the intra-frame coding mode (intra coding), and is set on the terminal 5B side in the forward prediction coding mode. Sometimes it is set to the terminal 5C side. Even in the case of a P picture or a B picture, if the code amount is reduced in the intra coding, the intra frame coding mode is set.
[0027]
First, a case where an I picture is transmitted will be described. When transmitting an I picture, intra-frame predictive coding is performed. At this time, the switch circuit 5 is set to the terminal 5A side. The frame memory 2 outputs image data of the current frame. The image data of the current frame is supplied to the DCT circuit 6 via the switch circuit 5.
[0028]
The DCT circuit 6 converts (8 × 8) into one block, and the time-domain video signal is DCT-converted and converted into spectrum data in the frequency domain. This spectrum data is read out in order from the DC component by so-called zigzag scanning. The output of the DCT conversion circuit 6 is supplied to the quantization circuit 7.
[0029]
The quantization value is supplied to the quantization circuit 7 from the quantization control circuit 22. Information on the output bit rate is supplied from the transmission buffer 9 to the quantization control circuit 22.
[0030]
In the quantization circuit 7, the spectrum data from the DCT changing circuit 6 is quantized according to the quantization value from the quantization control circuit 22. Thereby, the bit rate of the output bit stream is kept constant.
[0031]
The output of the quantization circuit 7 is supplied to the variable length coding circuit 8 and also to the inverse quantization circuit 10. In the variable length encoding circuit 8, the spectrum data from the quantization circuit 7 is variable length encoded together with the motion vector, the quantization scale, and the prediction mode.
[0032]
From the variable length coding circuit 8, the coded bit stream is output. This bit stream is temporarily stored in the transmission buffer 9. A data stream is read from the transmission buffer 9 according to the required bit rate, and a bit stream is output from the output terminal 11.
[0033]
Further, the output of the quantization circuit 7 is supplied to the inverse quantization circuit 10. The output of the inverse quantization circuit 10 is supplied to the IDCT circuit 12. In the case of an I picture, since the DCT conversion is performed on the pixels in the frame, the original screen is formed by the inverse quantization circuit 10 and the IDCT circuit 12. The digital video data for one screen is supplied to the frame memory 14 via the adding circuit 13. The data stored in the frame memory 14 is used as reference frame data in the next P picture or B picture.
[0034]
Next, a case where a P picture is transmitted will be described. When transmitting a P picture, forward predictive coding is performed with respect to the reference frame. At this time, the switch circuit 5 is set to the terminal 5B side. The frame memory 2 outputs a digital video signal of the current frame. The output of the frame memory 2 is supplied to the subtraction circuit 6.
[0035]
The frame memory 14 stores reference frame data. The reference frame data is motion-compensated by the motion compensation circuit 15 and supplied to the subtraction circuit 6. The subtraction circuit 6 obtains the difference between the current frame data and the motion compensated reference frame data. Difference data between the current frame data and the reference frame data is supplied to the DCT circuit 6 via the switch circuit 5.
[0036]
The DCT circuit 6 performs DCT conversion on the difference data between the reference frame data and the current frame data, and converts it into spectrum data. The output of the DCT conversion circuit 6 is supplied to the quantization circuit 7. The quantizing circuit 7 quantizes the spectrum data from the DCT changing circuit 6.
[0037]
The quantization value is supplied to the quantization circuit 7 from the quantization control circuit 22. Information on the output bit rate is supplied from the transmission buffer 9 to the quantization control circuit 22. Further, the output of the termination macroblock detection circuit 21 is supplied to the quantization control circuit 22.
[0038]
In the quantization circuit 7, the spectrum data from the DCT changing circuit 6 is quantized according to the quantization value from the quantization control circuit 22. Thereby, the bit rate of the output bit stream is kept constant. Also, the quantization value is lowered in the macroblock at the end of the screen. As a result, image quality degradation due to a decrease in the reliability of the motion vector is compensated.
[0039]
The output of the quantization circuit 7 is supplied to the variable length coding circuit 8 and also to the inverse quantization circuit 10. In the variable length encoding circuit 8, the spectrum data from the quantization circuit 7 is variable length encoded together with the motion vector, the quantization scale, and the prediction mode.
[0040]
From the variable length coding circuit 8, the coded bit stream is output. This bit stream is temporarily stored in the transmission buffer 9. A data stream is read from the transmission buffer 9 according to the required bit rate, and a bit stream is output from the output terminal 11.
[0041]
Further, the output of the quantization circuit 7 is supplied to the inverse quantization circuit 10. The output of the inverse quantization circuit 10 is supplied to the IDCT circuit 12. In the case of a P picture, the difference between the reference frame data and the current screen data is DCT transformed, so that the difference between the reference frame data and the current screen data is calculated by the inverse quantization circuit 10 and the IDCT circuit 12. can get. This difference data is supplied to the addition circuit 13.
[0042]
The reference screen data up to that point is supplied from the frame memory 14 to the adder circuit 13 via the motion compensation circuit 15. The adding circuit 13 adds the difference data to the data of the reference screen so far. The output of the adder circuit 13 is accumulated in the frame memory 14 as data of the next reference frame.
[0043]
Next, a case where a B picture is transmitted will be described. When transmitting a B picture, bidirectional prediction encoding is performed using a past reference frame and a future reference frame. At this time, the switch circuit 5 is set to the terminal 5C side. The frame memory 2 outputs a digital video signal of the current frame. The output of the buffer memory 2 is supplied to the subtraction circuit 7.
[0044]
The frame memory 4 stores data of past reference frames and future reference frames. The data of both reference frames are motion compensated by the motion compensation circuit 5 and supplied to the subtraction circuit 7. The subtraction circuit 7 obtains the difference between the current frame data and the motion compensated past and future reference frame data. Difference data between the current frame data and past and future reference frame data is supplied to the DCT circuit 6 via the switch circuit 5.
[0045]
The DCT circuit 6 performs DCT conversion on the difference data between the data of the past and future reference frames and the data of the current frame, and converts the data into spectrum data. The output of the DCT conversion circuit 6 is supplied to the quantization circuit 7. The quantizing circuit 7 quantizes the spectrum data from the DCT changing circuit 6.
[0046]
The quantization value is supplied to the quantization circuit 7 from the quantization control circuit 22. Information on the output bit rate is supplied from the transmission buffer 9 to the quantization control circuit 22. Further, the output of the termination macroblock detection circuit 21 is supplied to the quantization control circuit 22.
[0047]
In the quantization circuit 7, the spectrum data from the DCT changing circuit 6 is quantized according to the quantization value from the quantization control circuit 22. Thereby, the bit rate of the output bit stream is kept constant. Also, the quantization value is lowered in the macroblock at the end of the screen. As a result, image quality degradation due to a decrease in the reliability of the motion vector is compensated.
[0048]
The output of the quantization circuit 7 is supplied to the variable length coding circuit 8 and also to the inverse quantization circuit 10. In the variable length encoding circuit 8, the spectrum data from the quantization circuit 7 is variable length encoded together with the motion vector, the quantization scale, and the prediction mode.
[0049]
From the variable length coding circuit 8, the coded bit stream is output. This bit stream is temporarily stored in the transmission buffer 9. A data stream is read from the transmission buffer 9 according to the required bit rate, and a bit stream is output from the output terminal 11.
[0050]
Thus, in the example of the MPEG2 system encoder to which the present invention is applied, the termination macroblock detection circuit 21 is provided to detect whether or not a motion vector is detected in the macroblock at the end of the screen. The quantization value for the quantization circuit 7 is switched between when the macroblock at the end of the screen is detected and when it is not. As a result, it is possible to prevent the deterioration of the image quality in a place where the motion vector near the end of the screen has low reliability.
[0051]
That is, as shown in FIG. 2A, when a macro block near the center of the screen is detected, the search range SA is within the screen. For this reason, when the block BLK1 and the block BLK2 are matching blocks, the matching block is correctly detected and the motion vector V1 is correctly obtained.
[0052]
On the other hand, as shown in FIG. 2B, when a macroblock near the end of the screen is detected, the search range SA is outside the screen. For this reason, even if the block BLK3 and the block BLK4 are matching blocks, the correct motion vector V2 cannot be detected because the block BLK3 comes outside the screen.
[0053]
Thus, the reliability of the motion vector in the macro block near the end of the screen is lower than the reliability of the motion vector in the macro block near the center of the screen. For this reason, image quality degradation occurs near the end of the screen.
[0054]
Therefore, in this example, when encoding is performed using a B picture or a P picture, the quantization value is reduced at the end of the macroblock. That is, FIG. 3 shows a macroblock of one screen. In the macroblock of one screen, in the area AR1 at the end of the macroblock as shown by diagonal lines, the quantization control circuit 22 is set so that the quantum value becomes small. Be controlled. This compensates for the deterioration in image quality due to a decrease in the reliability of the motion vector.
[0055]
FIG. 4 is a flowchart showing the quantization process at this time. In FIG. 4, the number of target bits is determined for each picture type (step S1). A quantization value Qj for the j-th macroblock being processed is determined (step S2). It is determined whether or not the j-th macroblock being processed is located at the end of the screen (step S3).
[0056]
If the j-th macroblock is not located at the end of the screen, the quantization value is not changed, and the quantization process is performed with the quantization value Qj determined in step S2 (step S4).
[0057]
If the j-th macroblock is located at the end of the screen, for example, 0.5 is multiplied to the quantized value Qj determined in step S2 (step S5). Quantization processing is performed with this 0.5Qj (step S4).
[0058]
When the quantization process is performed in step S4, it is determined whether or not a macro block for one frame has been processed (step S6). If the macro block for one frame is not processed, the process returns to step S2 and the above-described processing is repeated. When the macroblock for one frame is processed, the quantization process is terminated.
[0059]
Thus, in this example, in the macroblock at the end of the screen, the quantized value Qj is multiplied by 0.5 to reduce the quantized value. In this way, since the quantized value of the macroblock at the end of the screen is reduced, the deterioration of the image quality due to the decrease in the reliability of the motion vector is compensated.
[0060]
By the way, even in the case of a P picture or a B picture, when the code amount is reduced by encoding within a frame, an intra-frame encoding process (intra encoding) is performed. When intra processing is performed, since no motion vector is required, even when the j-th macroblock is at the end of the screen, the image quality does not deteriorate, and there is no need to change the quantization value.
[0061]
FIG. 5 is a flowchart showing an example in consideration of the case where such intra processing is performed. In FIG. 5, the number of target bits is determined for each picture type (step S11). A quantization value Qj for the j-th macroblock being processed is determined (step S12).
[0062]
It is determined whether or not the j-th macroblock is a macroblock for intra coding processing (step S13). If the j-th macroblock is a macroblock for intra coding processing, the quantization value is not changed, and the quantization processing is performed with the quantization value Qj determined in step S12 (step S14).
[0063]
If it is determined in step S13 that the jth macroblock is not a macroblock for intra coding processing, it is determined whether the jth macroblock is a macroblock at the end of the screen (step S15).
[0064]
If the j-th macroblock is not the macroblock at the end of the screen, the quantization value is not changed, and the quantization process is performed with the quantization value Qj determined in step S12 (step S14).
[0065]
If the j-th macroblock is the macroblock at the end of the screen, for example, 0.5 is multiplied to the quantized value Qj determined in step S12 (step S16). With this 0.5Qj, the quantization process is performed (step S14).
[0066]
When the quantization process is performed in step S14, it is determined whether or not a macro block for one frame has been processed (step S17). If the macroblock for one frame is not processed, the process returns to step S12 and the above-described processing is repeated. When the macroblock for one frame is processed, the quantization process is terminated.
[0067]
FIG. 6 shows another example of a quantization value setting process for compensating for image quality deterioration due to a decrease in the reliability of the motion vector.
In the above-described example, the quantum value is set to 0.5 in the area AR1 at the end of the macroblock in FIG. 3 so as to prevent image quality deterioration due to a decrease in the reliability of the motion vector. In this example, Furthermore, the terminal area AR11 and the second area AR12 from the terminal end shown in FIG. 7 are detected. When the macro block is detected in the second area AR12 from the end, the quantized value is decreased. When the macro block is detected in the end area AR11, the quantized value is further decreased. I have to.
[0068]
In FIG. 6, the number of target bits is determined for each picture type (step S21). A quantization value Qj for the j-th macroblock being processed is determined (step S22).
[0069]
It is determined whether or not the j-th macroblock being processed is located within 3 macroblocks from the edge of the screen (step S23). If the jth macroblock being processed is within 3 macroblocks from the edge of the screen, it is determined whether or not it is positioned at the end of the screen (step S24). If it is not located at the end of the screen, it is determined whether or not it is located at the second macroblock from the end of the screen (step S25).
[0070]
If it is determined in step S23 that the j-th macroblock being processed is not located within 3 macroblocks from the end of the screen, the quantization value is not changed, and the quantization determined in step S22 is performed. Quantization processing is performed with the value Qj (step S26).
[0071]
If it is determined in step S24 that the j-th macroblock being processed is located at the end of the screen, the quantized value Qj determined in step S22 is multiplied by, for example, 0.5 ( Step S27). In step S26, a quantization process is performed with the changed quantization value 0.5Qj.
[0072]
If it is determined in step S25 that the j-th macroblock being processed is located at the second macroblock from the end of the screen, the quantization value Qj determined in step S22 is, for example, 0. 75 is multiplied (step S28). In step S26, quantization processing is performed with the changed quantization value 0.75Qj.
[0073]
If it is determined in step S25 that the j-th macroblock being processed is not located at the second macroblock from the end of the screen, the quantization value Qj determined in step S22 is, for example, 0. 9 is multiplied (step S29). In step S26, a quantization process is performed with the changed quantization value 0.9Qj.
[0074]
When the quantization process is performed in step S26, it is determined whether or not a macro block for one frame has been processed (step S30). If the macro block for one frame is not processed, the process returns to step S22 and the above-described processing is repeated. When the macroblock for one frame is processed, the quantization process is terminated.
[0075]
In the above example, it is detected whether or not it is at the position of the macroblock at the end of the screen and whether or not it is at the position of the second macroblock from the end of the screen. The quantized value may be switched by detecting whether or not it is in the fourth, fourth,... Macroblock.
[0076]
Of course, as in the example shown in FIG. 4, if the j-th macroblock is to be intra-processed, the quantized value is used regardless of whether it is the second from the end of the screen or the end. May not be changed.
[0077]
FIG. 8 shows still another example of the quantization value setting process for compensating for the image quality deterioration due to the decrease in the reliability of the motion vector.
In the above examples, the macro block near the end of the screen has low motion vector reliability. However, the macro block near the end of the screen has a smaller quantized value, resulting in a decrease in motion vector reliability. The image quality is compensated for.
[0078]
However, it is not only in the case of the macroblock at the end of the screen that the reliability of the motion vector decreases. Even when the input image is difficult, that is, when the pattern is complex and the motion is intense, the reliability of the motion vector is lowered.
[0079]
Therefore, in this example, when the difficulty of the input image is large, the quantization value is lowered to compensate for the image quality deterioration due to the decrease in the reliability of the motion vector.
[0080]
In this example, the sum of the most recent P picture encoding average quantization value and the B picture encoding average quantization value is used as an index as to whether or not the difficulty of the input image is large.
[0081]
In FIG. 8, the number of target bits is determined for each picture type (step S31). The quantization value Qj for the jth macroblock being processed is determined (step S32).
[0082]
It is determined whether or not the sum of the average quantization value Qp at the time of encoding the latest P picture and the average quantization value Qb at the time of encoding the latest B picture is smaller than 60 (step S33). If the sum of the average quantization value Qp at the time of encoding the latest P picture and the average quantization value Qb at the time of encoding the latest B picture is smaller than 60, for example, the difficulty of the input image is low and the reliability of the motion vector is high. The quantization value is not changed, and the quantization process is performed with the quantization value Qj determined in step S32 (step S34).
[0083]
In step S33, if the sum of the average quantization value Qp for the latest P picture encoding and the average quantization value Qb for the latest B picture encoding is not smaller than 60, for example, the average quantization value Qb for the latest B picture encoding It is determined whether or not the sum is less than 80, for example (step S35).
[0084]
If the sum of the average quantization value Qp at the time of the latest P picture encoding and the average quantization value Qb at the time of the latest B picture encoding is smaller than 80, for example, for the quantization value Qj determined in step S32, for example, 0.75 is multiplied (step S36). Quantization processing is performed at 0.75Qj (step S34).
[0085]
In step S35, if the sum of the average quantization value Qp for the latest P picture encoding and the average quantization value Qb for the latest B picture encoding is not smaller than 80, for example, the quantization value Qj determined in step S32 is set. On the other hand, for example, 0.5 is multiplied (step S37). With this 0.5Qj, a quantization process is performed.
[0086]
When the quantization process is performed in step S34, it is determined whether or not a macro block for one frame has been processed (step S38). If the macroblock for one frame is not processed, the process returns to step S32 and the above-described processing is repeated. When the macroblock for one frame is processed, the quantization process is terminated.
[0087]
Thus, in this example, when the sum of the average quantization value Qp at the time of the latest P picture encoding and the average quantization value Qb at the time of the latest B picture encoding is larger than 60, for example, the input image is difficult, and step S32 For example, 0.75 is multiplied by the quantized value Qj determined in (5), and the sum of the average quantized value Qp at the time of encoding the latest P picture and the average quantized value Qb at the time of encoding the latest B picture is 80, for example. When it is larger, it is further difficult to multiply the quantized value Qj determined in step S12 by, for example, 0.5. As a result, image quality degradation due to a decrease in the reliability of the motion vector is compensated.
[0088]
In the above-described example, the sum of the average quantization value Qp at the time of the latest P picture encoding and the average quantization value Qb at the time of the latest B picture encoding is used as an index of difficulty of the input image. Can also be used. For example, the average quantization value Qp when encoding the latest P picture or the average quantization value Qb when encoding the latest B picture may be used as an index of difficulty of the input image.
[0089]
Further, in this example, two thresholds are set for the sum of the average quantization value Qp at the time of nearest P picture coding and the average quantization value Qb at the time of nearest B picture coding used as an index of difficulty of the input image. The quantized value is set by using a single threshold value, or a plurality of threshold values may be set.
[0090]
Of course, as in the example shown in FIG. 4, if the j-th macroblock is to be intra-coded, the average quantization value Qp at the time of the latest P picture coding and the average at the time of the latest B picture coding Regardless of the sum of the quantization value Qb, the quantization value may not be changed.
[0091]
Alternatively, the quantization value may be determined by combining information indicating whether there is a macroblock near the end of the screen and information on the difficulty of the input image. That is, the quantized value may be set by combining the process shown in FIG. 4 or 5 and the process shown in FIG.
[0092]
In the above example, the MPEG2 system encoding process has been described. However, the present invention is not limited to the MPEG2 system, and can be similarly applied to the MPEG1 system and other systems that perform motion compensation predictive coding.
[0093]
【The invention's effect】
According to the present invention, in the macro block near the end of the screen, the quantization value is lowered and the code amount is increased in the portion where the reliability of the motion vector is considered to be lowered. This compensates for image quality degradation in a portion where the reliability of the motion vector near the end of the screen is reduced.
[0094]
In addition, according to the present invention, the quantization value is lowered in a portion where the most recent average quantization value is equal to or greater than a predetermined value, the image encoding is difficult, and the reliability of the motion vector is considered to be reduced. The code amount is increased. This compensates for image quality degradation in a portion where the reliability of the motion vector near the end of the screen is reduced.
[Brief description of the drawings]
FIG. 1 is a block diagram of an example of a motion MPEG2 encoder to which the present invention is applied.
FIG. 2 is a schematic diagram for explaining the reliability of a motion vehicle.
FIG. 3 is a schematic diagram used for explaining an example of a motion MPEG2 encoder to which the present invention is applied;
FIG. 4 is a flowchart used for explaining an example of a motion MPEG2 encoder to which the present invention is applied;
FIG. 5 is a flowchart used to describe another example of a motion MPEG2 encoder to which the present invention is applied;
FIG. 6 is a flowchart used to describe still another example of the motion MPEG2 encoder to which the present invention is applied;
FIG. 7 is a schematic diagram used for explaining still another example of the motion MPEG2 encoder to which the present invention is applied;
FIG. 8 is a flowchart used to describe still another example of the motion MPEG2 encoder to which the present invention is applied;
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 3 ... Motion vector detection circuit, 7 ... Quantization circuit, 21 ... Terminal macroblock detection circuit, 22 ... Quantization control circuit

Claims (7)

Motion vector detecting means for obtaining a motion vector between the input image and the reference image;
Orthogonal transform means for performing motion compensation based on the motion vector obtained by the motion vector detection means and orthogonally transforming the difference between the reference image and the input image;
Quantization means for quantizing the output of the orthogonal transform means;
When the difficulty of the input image exceeds a predetermined value, it is determined that the reliability of the motion vector detected by the motion vector detection unit is low, and a quantization value control unit that decreases the quantization value of the quantization unit; With
The difficulty of the said input image is the encoding apparatus which is the average quantization value at the time of the last encoding. The quantized value control means reduces the quantized value of the quantizing means for a macroblock located near the end of the input image, assuming that the motion vector has low reliability. Encoding device. The quantized value control means divides the distance from the end of the input image into a plurality, and changes the quantized value of the quantizing means stepwise according to the distance from the end of the input image. 2. The encoding device according to 2. The encoding apparatus according to claim 1, wherein the quantized value control means does not change the quantized value for the macroblock that performs the intra-picture encoding, regardless of the reliability of the motion vector.   The encoding apparatus according to claim 1, wherein the difficulty of the input image is a sum of an average quantization value at the time of the latest P picture encoding and an average quantization value at the time of the latest B picture encoding. The quantized value control means compares an index indicating difficulty in encoding the input image with a plurality of threshold values, and changes the quantized value of the quantizing means stepwise according to the threshold values. The encoding device according to 1. Find the motion vector between the input image and the reference image,
Performing motion compensation based on the obtained motion vector to orthogonally transform the difference between the reference image and the input image,
When the difficulty of the input image exceeds a predetermined value, the reliability of the obtained motion vector is low, the quantization value is reduced, and the orthogonally transformed output is quantized.
The difficulty of the said input image is the encoding method which is the average quantization value at the time of the last encoding.

Images