JP2005328383A - Dynamic image encoding device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a dynamic image encoding device which can select a predictive image close to an actual image to be encoded, while suppressing an increase in the amount of operation on motion compensation. <P>SOLUTION: The dynamic image encoding device comprises a motion vector search range specifying part 12 for specifying a search range of the motion vector; a predictive image area specifier 13 for specifying a predictive image specifier to be compensated for motion prediction within a search range of the motion vector; an image corrector 2 for correcting an image characteristic value of the predictive image so as to match an image characteristic value of an image to be encoded as correction for motion prediction; an evaluation computer 10 for computing a predictive evaluation value of the predictive image on the basis of matching processing of the predictive image corrected for the motion prediction and the image to be encoded; and a motion vector selecting 11 for selecting a motion vector used for encoding the image to be encoded based on the predictive evaluation value. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a moving image encoding apparatus that encodes a moving image by motion prediction, and a program for causing a computer to realize the same.

  As a conventional moving image encoding device, for example, there is one disclosed in Patent Document 1. In this apparatus, based on the average value of the pixel values in the encoding target frame or the encoding target block and the average value of the pixel values in the reference image frame or the predicted image block, the pixel values in the encoding target block are calculated. Correct the value. For example, correction is performed by subtracting the average value of each pixel in each block from the value of the pixel in the encoding target block and the value of the pixel in the predicted image block.

  In addition to this, Patent Document 1 discloses that the average value of the pixels in the encoding target frame (or the encoding target block) is changed to the average value of the pixels in the reference image frame (or the predicted image block). Processing for correcting each pixel in the encoding target block so as to match, the average value of the pixel value in the encoding target block is equal to the average value of the pixel value in the reference image frame Processing for correcting the value of each pixel in the encoding target block, or according to the ratio of the average value of the pixel values in the encoding target frame and the average value of the pixel values in the reference image frame A process for correcting the value is disclosed.

  Based on the difference between the pixel value in the coding target block corrected as described above and the pixel value in the predicted image block, a prediction evaluation value such as a sum of absolute differences of pixel values is calculated, and this prediction evaluation is performed. Select the motion vector with the smallest value. Then, the encoding target block is encoded in accordance with the encoding mode determined based on the prediction evaluation value.

  By applying such processing to the block to be encoded, even if the brightness of the entire screen changes over time, the prediction evaluation value is obtained while removing the influence of the brightness change and moves. Running compensation.

JP 2000-106675 A

  In the conventional moving image encoding device, when obtaining the average value of the pixel values in the encoding target block and the prediction image block, it is necessary to perform an operation using the average value for each block accordingly. There has been a problem that the computation time related to compensation prediction is prolonged, and hardware such as a memory for storing data required for the computation is also required.

  Also, the pixels in the encoding target block so that the average value of the pixels in the encoding target frame (or encoding target block) matches the average value of the pixels in the reference image frame (or prediction image block). The process of correcting the value deviates from the original purpose of motion search, which is to search for the predicted image closest to the encoding target image, and can search for the predicted image closest to the encoding target image. Is not limited.

  For example, in motion search, the reference image frame and the encoding target block are generally not the same size, and the motion vector search region is often a partial region of the reference image frame. For this reason, even if correction is performed so that the average value of the pixel values in the encoding target block matches the average value of the pixel values in the reference image frame, it is actually a motion search target in the reference image frame. There may be a case where the average value of the pixels in the partial area does not match.

  This is because the pixel value in the encoding target block is corrected so that the average value of the pixel value in the encoding target frame is equal to the average value of the pixel value in the reference image frame, or the reference image frame The same applies to the case where the pixel value in the encoding target block is corrected in accordance with the ratio of the average value of the pixel values in the encoding target frame and the average value of the pixel values in the encoding target frame.

  Moreover, in the moving image encoding device disclosed in Patent Document 1, the prediction evaluation value calculated based on the difference between the pixel value in the corrected encoding target block and the pixel value in the corrected prediction image block is Although the minimum motion vector is selected, the encoding process itself is executed by inputting the original image to be encoded and the original predicted image that have not been corrected. For this reason, there is a possibility that there is a motion vector with good encoding efficiency in addition to the motion vector having the smallest predicted evaluation value calculated based on the image in which the pixel value is corrected, and this cannot be fully searched.

  For example, in MPEG-2, a variable-length code is assigned to a DC component coefficient (DC component coefficient) in inter-frame prediction, which is inter prediction equivalent to forward prediction, backward prediction, bidirectional prediction, and the like. The variable length encoded code is assigned a code length proportional to the size of the DC component coefficient. For example, the DC component coefficient obtained from the difference between the original reference image whose pixel value is not corrected and the encoding target image The longer the variable length coding code is assigned, the larger.

  Accordingly, the DC component coefficient obtained from the predicted image predicted with the motion vector that minimizes the predicted evaluation value calculated based on the image with the corrected pixel value is, for example, the next smaller than the one with the smallest predicted evaluation value (predictive evaluation). If the value of the DC component coefficient obtained by the predicted image predicted by the motion vector (the second smallest value) is larger than the value of the DC component coefficient, the predicted image corresponding to the motion vector having the second smallest estimated evaluation value is encoded. High efficiency.

  The present invention has been made to solve the above-described problems, and selects a motion vector based on a matching evaluation value obtained by correcting a pixel value of a predicted image region within a motion vector search range. By doing so, it is possible to select a predicted image that is close to the actual encoding target image while suppressing an increase in the amount of computation related to motion compensation. For example, the luminance of a part of the frame changes like a flash image. To obtain a moving picture coding apparatus capable of selecting a motion vector with high coding efficiency even if the moving picture and the picture including the image are the same, and the luminance and color difference are different, and a program for causing the computer to realize the moving picture coding apparatus With the goal.

  A moving image encoding apparatus according to the present invention includes a search range specifying unit that specifies a search range of a motion vector, and an area specifying unit that specifies a predicted image specifying region to be corrected for motion prediction within the search range of a motion vector A prediction image correction unit that corrects the image feature value of the prediction image in the prediction image designation region so as to match the image feature value of the encoding target image as a motion prediction correction, and a prediction subjected to the motion prediction correction An evaluation value calculation unit that calculates a prediction evaluation value of a prediction image indicating a degree of difference between the image and the encoding target image by a matching process between the image and the encoding target image, and a code of the encoding target image based on the prediction evaluation value And a motion vector selection unit that selects a motion vector used for conversion.

  According to the present invention, a search range designating unit for designating a motion vector search range, a region designating unit for designating a predicted image designating region to be corrected for motion prediction within the motion vector search range, and a motion prediction A prediction image correction unit that corrects the image feature value of the prediction image in the prediction image designation region so as to match the image feature value of the encoding target image, and a prediction image that has undergone correction for motion prediction and the encoding target An evaluation value calculation unit that calculates a prediction evaluation value of a prediction image indicating a degree of difference from the encoding target image by matching processing with the image, and a motion vector used for encoding the encoding target image based on the prediction evaluation value A motion vector selection unit that selects the image to be corrected so that the predicted image of interest specified within the range in which the motion vector search is actually performed matches the image feature value of the encoding target image. From Rukoto, there is an effect that it is possible to select an image close to the pattern of the input image. As a result, for example, images with brightness changes such as fade-in, fade-out, and flash images, or images with similar patterns but different brightness and color differences, etc. It is possible to select a prediction image with high encoding efficiency even if a motion vector that does not accurately follow the motion of the encoding target image is selected and the correct motion vector is likely to be excluded from the selection candidates. effective.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a configuration of a moving picture coding apparatus according to Embodiment 1 of the present invention. The moving image encoding apparatus 1 according to Embodiment 1 includes an image correction unit 2, a motion prediction unit 3, an image memory unit 4, a difference calculation unit 5, an encoding unit 6, a decoding unit 7, a decoding addition unit 8, and a variable length code. It comprises the conversion unit 9. The moving image encoding apparatus 1 can be realized by executing a moving image encoding program according to the present invention using, for example, a general-purpose computer.

  That is, by causing the computer to execute the moving image encoding program to function as the constituent elements 2 to 9, characteristic data processing in the moving image encoding device of the present invention can be executed. In the following description, the configuration of the computer itself that embodies the video encoding device of the present invention and the basic functions thereof can be easily recognized by those skilled in the art based on the common general technical knowledge in the field, Since it is not directly related to the essence of the present invention, detailed description will be omitted.

  FIG. 2 is a block diagram showing a configuration of the motion prediction unit in FIG. The motion prediction unit 3 includes an evaluation value calculation unit 10, a motion vector selection unit 11, a motion vector search range designation unit (search range designation unit) 12, and a predicted image region designation unit (region designation unit) 13. In the following description, a configuration in which the motion vector search range specifying unit 12 and the predicted image region specifying unit 13 are provided inside the motion predicting unit 3 is shown as an example. However, the motion vector search range specifying unit 12 and the predicted image region specifying Either one or both of the units 13 may be configured as elements independent of the motion prediction unit 3.

  Also in this case, by configuring the motion vector search range specifying unit 12 and the predicted image region specifying unit 13 as one program module of the moving image encoding program, the constituent elements of the moving image encoding device of the present invention As a computer can function.

Next, the operation will be described.
In the present embodiment, a case where each frame image is divided into a plurality of image blocks and encoded for each block will be described as an example. Note that the present invention is not limited by the shape and block size of the image block.

  First, an encoding target block obtained by dividing a frame image of a moving image to be encoded is input to the image correction unit 2, the motion prediction unit 3, and the difference calculation unit 5 as input image data a. The input image data a includes data for specifying an image block position in the frame image of the encoding target image. The input image data a input to the motion prediction unit 3 is further input to the internal evaluation value calculation unit 10 and the motion vector search range specification unit 12.

  The motion vector search range specifying unit 12 specifies an image block position in the encoding target frame image from the input image data a, and searches for a motion vector corresponding to the encoding target block according to the specified position. Is specified. As a method for specifying the motion vector search range, an appropriate range in the reference image may be determined according to the position of the encoding target block. Information for specifying the motion vector search range specified in this way is generated by the motion vector search range specifying unit 12 and output to the predicted image region specifying unit 13.

  The predicted image region designating unit 13 designates a part or all of the region within the motion vector search range designated by the motion vector search range designating unit 12 as a predicted image designated region, and reads the predicted image data of the designated region. Is generated and output to the image memory unit 4.

The image memory unit 4 stores image data of a reference image frame that is referred to when motion prediction is performed on the encoding target block. Based on the predicted image designation area read signal d input from the predicted image area designation section 13, the image memory section 4 designates a reference image frame with a motion vector search range and a predicted image designation area within the range. As a predicted image frame, the predicted image data f in the predicted image designation area is read from the predicted image frame data and output to the predicted image correction unit 2.
In the above example, the input image data a includes data such as position information for specifying the image block position in the frame image of the encoding target image, but the data such as the position information is input image data. It may be other data independent of a, and does not limit the present invention.

FIG. 3 is a diagram showing a predicted image frame, a motion search range defined in the predicted image frame, a predicted image designation area, and a predicted image block stored in the image memory unit in FIG. As shown in the figure, the motion vector search range 15 is specified by the motion vector search range specifying unit 12 in the predicted image frame 14 stored in the image memory unit 4.
The illustrated example shows a case where the motion vector search range 15 is an area smaller than the predicted image frame 14, but the motion vector search range only needs to be the same as or included in the predicted image frame 14. The present invention is not limited by the size of 15.

  In the illustrated example, a predicted image specifying area 16 is specified by the predicted image area specifying unit 13 as a partial area in the search range 15. In the present invention, the predicted image block 17 indicated by the hatched area is extracted from the predicted image designation area 16 and used for motion prediction. The predicted image block 17 has the same size as the input image block to be encoded.

  In the illustrated example, the predicted image designation region 16 is a region smaller than the motion vector search range 15 and a region larger than the predicted image block 17, but is not limited to this magnitude relationship. .

  For example, the predicted image designation area 16 may be the same size as the motion vector search range 15 or the same size as the predicted image block 17. In other words, it may be an area that is the same as or included in the motion vector search range 15 and that is the same size or larger than the predicted image block 17. This is because a prediction image block corresponding to the input image block to be encoded is selected from the motion vector search range 15.

  When the input image data a of the input image block to be encoded and the predicted image data f of the predicted image designation area 16 are input to the image correction unit 2, the predicted image designation area 16 used for motion prediction based on these data a and f. The predicted image is corrected.

Here, prediction image correction processing by the image correction unit 2 will be described in detail.
FIG. 4 is a block diagram showing the configuration of the image correction unit in FIG. As shown in the figure, the image correction unit 2 includes an input image feature value calculation unit 18, a predicted image feature value calculation unit 19, and a predicted image correction unit 20.

  The input image feature value calculation unit 18, the predicted image feature value calculation unit 19, and the predicted image correction unit 20 are configured as a program module that configures a moving image encoding program according to the present invention. As a result, a computer that executes the moving picture encoding program can be caused to function as the constituent elements 18 to 20.

  The input image feature value calculation unit 18 inputs the input image data a of the input image block to be encoded, analyzes the image feature, and calculates an input image feature value obtained by digitizing the image feature. On the other hand, the predicted image feature value calculation unit 19 inputs the predicted image data f in the predicted image designation area 16, analyzes the image feature, and calculates a predicted image feature value j obtained by quantifying the image feature.

  Note that the image feature value includes a pixel average value, a pixel maximum value, a pixel minimum value, a pixel value median value, a pixel value maximum frequency value for the predicted image in the predicted image designation area 16 and the pixel values in the input image block, A pixel value range or the like is conceivable, but any feature value that can be adjusted to the image feature value of the input image block by the addition / subtraction operation on the predicted image in the predicted image designation area 16 may be used.

  Further, as the image feature value, a feature value that best represents the image feature of the input image block is selected. The pixel value may be a pixel value for each of the color difference component, the RGB component, and other image signal components in addition to the luminance component.

  As a correction method, it is conceivable that the image feature value is corrected by multiplication using a weighting coefficient or the like. However, in the present invention, the prediction image in the prediction image designation region is corrected in the input image block for the following reason. Correction to match the image feature value is executed by addition / subtraction calculation.

  For example, a general process for DCT conversion of a prediction difference image in MPEG-2 or the like performs DCT conversion of a prediction difference image using an original prediction image before correction corresponding to a motion vector selected from the prediction image corrected by multiplication. . For this reason, the AC image (AC component) differs between the predicted image corrected by multiplication and the predicted image before correction, and the pattern of the predicted image may be different before and after correction.

  Therefore, by performing correction to adjust the image feature value of the predicted image to the image feature value of the input image block by addition / subtraction calculation, the DC component of the predicted image is different before and after the correction, but the AC component is not changed. In the present invention, since the purpose is to select a predicted image that most closely resembles the design of the input image block to be encoded, the predicted image is corrected by an addition / subtraction operation in which the AC component is not changed even by correction.

  The predicted image correction unit 20 receives the input image feature value from the input image feature value calculation unit 18, and further receives the predicted image data f and its predicted image feature value j from the predicted image feature value calculation unit 19, Based on the feature value and the predicted image feature value j, the predicted image defined by the predicted image data f is corrected. The predicted image data b of the corrected predicted image is output from the predicted image correction unit 20 to the motion prediction unit 3.

  Specifically, the predicted image feature value j of the predicted image designation area 16 designated for motion prediction within the motion vector search range 15 is matched with the input image feature value of the input image block obtained from the input image data a. The predicted image in the predicted image designation area 16 obtained from the predicted image data f is corrected.

  FIG. 5 is a graph showing the relationship between the predicted image in the predicted image designation area and the pixel value of the input image block to be encoded and the frequency of occurrence thereof. The predicted image feature value in the predicted image designation area is input to the input image data. A process for correcting the predicted image in the predicted image designation area so as to match the image feature value will be described. In the graphs shown in (a) to (f), the dashed curve is a curve representing the pixel value distribution of the predicted image in the predicted image designation area 16, and the solid curve is the pixel value of the input image block to be encoded. It is a curve showing distribution. The horizontal axis represents pixel values, and the vertical axis represents the frequency of occurrence of pixel values in each image area.

  (A) is a graph showing the relationship between the predicted image and the pixel value of the input image and the frequency of occurrence thereof, and a case where correction is performed to match the predicted image data with the image feature of the input image data using the pixel average value as the pixel feature value. Show. In the figure, a vertical broken line attached to a curve (broken line curve) representing the pixel value distribution of the predicted image in the predicted image designation area 16 and a vertical solid line attached to a curve (solid line curve) representing the pixel value distribution of the input image are as follows: The positions of the predicted image in the predicted image designation area 16 and the average pixel value of the input image block are shown.

  The predicted image correction unit 20 receives the image feature value (pixel average value) of the input image block from the input image feature value calculation unit 18 and further predicts the predicted image data f in the predicted image designation area 16 from the predicted image feature value calculation unit 19. When the image feature value j is input, the value of each pixel of the predicted image in the predicted image designation area 16 is increased (added) by the difference in pixel average value between the two images as indicated by the horizontal solid arrow in the figure. Thus, correction that matches the image feature value of the input image block is executed.

  (B) is a graph showing the relationship between the predicted image and the pixel value of the input image and the frequency of occurrence thereof, and a case where correction is performed to match the predicted image data to the image feature of the input image data using the pixel median as the pixel feature value. Show. In the figure, a vertical broken line attached to a curve (broken line curve) representing the pixel value distribution of the predicted image in the predicted image designation area 16 and a vertical solid line attached to a curve (solid line curve) representing the pixel value distribution of the input image are as follows: The positions of the predicted image in the predicted image designation area 16 and the pixel median value of the input image block are shown.

  The predicted image correction unit 20 receives the image feature value (pixel median value) of the input image block from the input image feature value calculation unit 18, and further predicts the predicted image data f in the predicted image designation area 16 from the predicted image feature value calculation unit 19. And its image feature value j are input, the value of each pixel of the predicted image in the predicted image designation area 16 is increased (added) by the difference in the pixel median value between the two images as shown by the horizontal solid arrow in the figure. Thus, correction that matches the image feature value of the input image block is executed.

  (C) is a graph showing the relationship between the predicted image and the pixel value of the input image and the frequency of occurrence thereof, and correcting the predicted image data to the image feature of the input image data using the maximum frequency value of the pixel value as the pixel feature value. Shows when to do. In the figure, a vertical broken line attached to a curve (broken line curve) representing the pixel value distribution of the predicted image in the predicted image designation area 16 and a vertical solid line attached to a curve (solid line curve) representing the pixel value distribution of the input image are as follows: The positions of the maximum frequency values of the pixel values of the predicted image and the input image block in the predicted image designation area 16 are shown.

  The predicted image correction unit 20 receives the image feature value (maximum frequency value of pixel values) of the input image block from the input image feature value calculation unit 18, and further predicts the predicted image designation area 16 from the predicted image feature value calculation unit 19. When the image data f and its image feature value j are input, the value of each pixel of the predicted image in the predicted image designation area 16 is increased by the difference in the maximum frequency value between the two images as indicated by the horizontal solid arrow in the figure ( Correction) to match the image feature value of the input image block.

  (D) is a graph showing the relationship between the predicted image and the pixel value of the input image and the frequency of occurrence, and correcting the predicted image data to match the image feature of the input image data with the maximum pixel value as the pixel feature value. Is shown. In the figure, a vertical broken line attached to a curve (broken line curve) representing the pixel value distribution of the predicted image in the predicted image designation area 16 and a vertical solid line attached to a curve (solid line curve) representing the pixel value distribution of the input image are as follows: The positions of the predicted image in the predicted image designation area 16 and the pixel value maximum value of the input image block are shown.

  The predicted image correction unit 20 receives the image feature value (pixel value maximum value) of the input image block from the input image feature value calculation unit 18 and further predicts the predicted image data in the predicted image designation area 16 from the predicted image feature value calculation unit 19. When f and its image feature value j are input, the value of each pixel of the predicted image in the predicted image designation area 16 is increased (added) by the difference between the maximum pixel values between the two images as indicated by the horizontal solid arrow in the figure. ) To perform correction to match the image feature value of the input image block.

  (E) is a graph showing the relationship between the predicted image and the pixel value of the input image and the frequency of occurrence, and correcting the predicted image data to match the image feature of the input image data with the minimum pixel value as the pixel feature value. Is shown. In the figure, a vertical broken line attached to a curve (broken line curve) representing the pixel value distribution of the predicted image in the predicted image designation area 16 and a vertical solid line attached to a curve (solid line curve) representing the pixel value distribution of the input image are as follows: The positions of the predicted image in the predicted image designation area 16 and the minimum pixel value of the input image block are shown.

  The predicted image correction unit 20 receives the image feature value (pixel value minimum value) of the input image block from the input image feature value calculation unit 18, and further predicts image data in the predicted image designation area 16 from the predicted image feature value calculation unit 19. When f and its image feature value j are input, the value of each pixel of the predicted image in the predicted image designation area 16 is increased (added) by the difference of the minimum pixel value between the two images as indicated by the horizontal solid arrow in the figure. ) To perform correction to match the image feature value of the input image block.

  (F) is a graph showing the relationship between the predicted image and the pixel value of the input image and the frequency of occurrence thereof, and correcting the predicted image data to match the image feature of the input image data with the pixel value range as the pixel feature value. Show. In the figure, a vertical broken line attached to a curve (broken line curve) representing the pixel value distribution of the predicted image in the predicted image designation area 16 and a vertical solid line attached to a curve (solid line curve) representing the pixel value distribution of the input image are as follows: Each pixel value range generated in the predicted image and the input image block in the predicted image designation area 16 is shown.

  The predicted image correction unit 20 receives the image feature value (pixel value range) of the input image block from the input image feature value calculation unit 18, and further predicts the predicted image data f in the predicted image designation region 16 from the predicted image feature value calculation unit 19. And its image feature value j are input, the value of each pixel of the predicted image in the predicted image designation area 16 is increased (added) by the difference in the pixel value range between the two images, as indicated by the horizontal solid arrow in the figure. Thus, correction that matches the image feature value of the input image block is executed.

  As described above, in the present invention, the image feature value of the predicted image in the predicted image designation area designated within the range in which the motion vector is actually searched is corrected to match the image feature value of the input image block.

  For example, in an image having a large luminance change such as a flash image, if a predicted evaluation value of a motion vector is calculated without correcting the predicted image, the predicted evaluation value tends to be too large. In this case, even a motion vector corresponding to the motion of the encoding target block may be excluded from motion vector selection candidates.

  On the other hand, when the predicted image is corrected as in the present invention, an image close to the input image block can be selected as a predicted image, and a predicted image with good coding efficiency can be selected even for a flash image. It becomes possible. In addition, an image that is most similar to the pattern of the input image block can be selected for predicted images that have similar patterns but differ in brightness, color difference, and the like.

  The evaluation value calculation unit 10 in the motion prediction unit 3 receives the input image data a and the prediction image data b from the prediction image correction unit 20, calculates a prediction evaluation value i, and outputs it to the motion vector selection unit 11. Specifically, using the predicted image data b, the predicted image block (the same as the predicted image specifying region 16) in the predicted image specifying region 16 corresponding to the input image block to be encoded specified by the input image data a. 17 (including the case of the size), and the corrected predicted image and the input image block to be encoded are matched, and the predicted image and the encoding target are compared using a pixel value difference between the images. A predicted evaluation value i (for example, sum of absolute value of pixel value difference) indicating the degree of difference from the input image block is calculated.

  As described above, in the present invention, since the predicted image is corrected by the addition / subtraction calculation, the corrected predicted image and the predicted image before correction have different direct current components (DC components). For this reason, the prediction image before correction with respect to the motion vector candidate with the smallest prediction evaluation value does not always have the best encoding efficiency. In addition, there may be a plurality of motion vectors having the smallest predicted evaluation value.

  Therefore, a threshold for evaluating the predicted evaluation value is set for the purpose of extracting some selection candidates for the predicted evaluation value in addition to the minimum predicted evaluation value. This threshold value may be set in advance as a fixed value, or may be a variable value that is arbitrarily set and updated from the image characteristics of the input image or predicted image. However, in any case, a value is set in a range in which motion vector candidates with good coding efficiency can be selected.

  For example, it is a value determined by taking statistics of parameters such as motion vector information amount, quantization step, S / N value, etc. for a plurality of different prediction evaluation values, and evaluating the image quality at the same time. Is set as the threshold value.

  The evaluation value calculation unit 10 performs comparison with the threshold value every time the predicted evaluation value i is calculated, excludes the predicted evaluation value i exceeding the threshold value, extracts a predicted evaluation value i less than the threshold value, and selects a motion vector To the unit 11.

  If there is one prediction evaluation value i extracted by the evaluation value calculation unit 10, the motion vector selection unit 11 selects a motion vector searched in the motion vector search range 15 based on the prediction evaluation value i. That is, the movement amount of the predicted image block corresponding to the predicted evaluation value i is obtained as a motion vector.

  On the other hand, when a plurality of prediction evaluation values i are extracted by the evaluation value calculation unit 10, the motion vector selection unit 11 considers encoding efficiency from among motion vector candidates determined by these prediction evaluation values i. Select a motion vector according to specified conditions.

  For example, for the predicted image corresponding to each of the extracted prediction evaluation values i, the image feature value of the original predicted image before the correction is determined to be closest to the image feature value of the input image block, and this image feature value A motion vector candidate is selected based on the predicted evaluation value i of the predicted image that gives. This is because the motion vector candidate closest to the image feature value of the input image block has the best DC component variable-length encoding efficiency.

  When there are a plurality of motion vector candidates closest to the image feature value of the input image block, the motion vector selection unit 11 performs the immediately preceding coded motion among the motion vector candidates corresponding to the plurality of prediction evaluation values described above. A motion vector with a small amount of information and good coding efficiency, such as a motion vector having the smallest sum of differences between the horizontal component and the vertical component, may be selected.

  Thus, in the present invention, instead of selecting only the smallest predicted evaluation value i, a plurality of predicted evaluation values i are extracted based on the threshold value, and a condition other than the predicted evaluation value (for example, for a predicted image) The motion vector corresponding to the plurality of prediction evaluation values is evaluated by a correction value or the like, and a motion vector having a high coding efficiency is selected. Thereby, an image close to the pattern of the input image can be selected. For example, it is possible to select a motion vector having the best variable length coding efficiency of DC component coefficients.

  When the motion vector selection unit 11 selects a motion vector as described above, the motion vector selection unit 11 generates motion vector information c regarding the selected motion vector and outputs the motion vector information c to the image memory unit 4 and the variable length encoding unit 9.

  In the image memory unit 4, the predicted image data e of the predicted image block corresponding to the motion vector information c is read and output to the difference calculation unit 5 and the decoding addition unit 8. The difference calculation unit 5 performs a difference calculation between the input image data a of the input image block to be encoded and the prediction image data e of the prediction image block corresponding to the input image block, and outputs the result to the encoding unit 6 as prediction difference image data. .

  The encoding unit 6 encodes the input prediction difference image data and outputs the encoded data to the decoding unit 7 and the variable length encoding unit 9. The decoding unit 7 decodes the encoded data from the encoding unit 6 and outputs the decoded data to the decoding addition unit 8 as predicted difference image decoded data. The encoding unit 6 and the decoding unit 7 may be arbitrary moving image encoding / decoding processes, and are not limited to MPEG-2, MPEG-4, vector quantization, or the like.

  The decoding addition unit 8 performs an addition operation on the predicted image data e and the prediction difference image decoded data from the decoding addition unit 8 and outputs the result to the image memory unit 4 as decoded addition image data g. This decoded added image data g becomes image data of a predicted image (reference) frame, and predicted image data f of the predicted image specified region in the predicted image frame is read from the image memory unit 4 based on the predicted image specified region read signal d. It is.

  In the variable length encoding unit 9, the motion vector information c and the encoded data from the encoding unit 6 are variable length encoded and output as variable length encoded data h. The variable length encoding process by the variable length encoding unit 9 may be an arbitrary variable length encoding process, and the contents of the present invention are not limited by the difference in the processing method.

  As described above, according to the first embodiment, the image feature of the prediction designated image is based on the prediction designated image designated within the range of the actual motion vector search and the image feature value of the input image block to be encoded. Since the value is corrected to match the image feature value of the input image block, the brightness change and brightness are similar to images with different brightness and images such as flash images but similar in brightness and color difference. Even if the motion vector that does not accurately follow the motion of the encoding target block due to differences in color or color difference is selected and the correct motion vector is likely to be excluded from the selection candidates, it is close to the pattern of the input image block An image can be selected. In addition, by encoding an image close to the design of the input image block, a prediction image with high encoding efficiency can be selected.

  In the first embodiment, the predicted image correction unit 20 converts the plurality of types of image feature values in the predicted image in the predicted image specification region 16 specified by the predicted image region specification unit 13 into the input image block to be encoded. In the above example, correction is performed so as to match the image feature values corresponding to each of the above-described types. A motion vector may be selected based on this.

  For example, a case where a pixel average value, a pixel value maximum frequency value, and a pixel value range are handled as a plurality of types of image feature values will be described. Of course, the image feature value is not limited to the pixel average value, the pixel value maximum frequency value, and the pixel value range, and is adjusted to the image feature value of the input image block by addition / subtraction operation on the prediction image of the prediction designated area. Any feature value may be used. For example, there are a median pixel value, a maximum pixel value, etc., and the type can be arbitrarily selected.

  First, the evaluation value calculation unit 10 obtains a prediction evaluation value by matching the prediction image corrected by the prediction image correction unit 20 based on each type of image feature value and the input image block to be encoded. Here, the predicted evaluation value is obtained using the predicted image corrected based on the pixel average value, the pixel value maximum frequency value, and the pixel value range.

  Subsequently, the motion vector selection unit 11 determines a motion vector based on the predicted evaluation value calculated for each image feature value by the evaluation value calculation unit 10. As a method of determining this motion vector, for example, a motion vector corresponding to the smallest one of prediction evaluation values calculated for each image feature value is selected, or each prediction evaluation value calculated for each image feature value is selected. A plurality of candidates are selected in ascending order, and a motion vector is determined by comprehensively determining a prediction evaluation value together with other factors such as a correction value of the prediction image (for example, a difference between image feature values of the prediction image and the input image). May be selected. Even with this configuration, the same effect as in the first embodiment can be obtained.

  In the first embodiment, the example in which the image feature value of the prediction image in the prediction designated region is corrected to match the image feature value of the input image block to be encoded has been described. In some cases, the pixel value in the corrected predicted image may exceed the range of pixel values that should be handled in advance in the moving picture coding apparatus according to the present embodiment.

  For example, in the moving picture encoding apparatus according to the present embodiment, consider a case in which it is specified in advance that a pixel value is expressed in an 8-bit range (of course, the present invention is not limited to 8 bits). 8 bits or less or 8 bits or more).

  In this case, as shown in FIG. 5A, when a pixel average value is selected as the image feature value and correction is performed to match the pixel average value of the predicted image in the prediction designated area with the pixel average value of the input image block, the correction should be performed. The combination of the average pixel value of the predicted image and the average pixel value of the input image block may exceed the 8-bit range described above. This tendency is the same even when the other image feature value described in the above embodiment is selected as the image feature value.

  Here, if the pixel value range is specified in advance, pixels in a further specified pixel value range may be set as pixels to be corrected by the predicted image. For example, the pixel value range of 16 to 235 out of the entire 8-bit range of 0 to 255 may be set as the effective region to be corrected even if the pixel value is not the entire pixel region expressed by 8 bits. In addition, when the pixel value range of the input image block is within the pixel value range designated in advance, the pixel value range of the input image block may be a pixel region to be corrected by the predicted image.

  Therefore, the predicted image correction unit 20 determines whether or not the pixel value of the predicted image corrected so as to match the image feature value of the input image block to be encoded is within a range that can be taken by the pixel value specified in advance. When there is a pixel value determined not to be within the pixel value range, the pixel value of the predicted image may be further corrected so as to be within a pixel value range designated in advance.

  As a correction method in this case, a lower limit or an upper limit boundary value is set in the pixel value range of the predicted image in the prediction specified region in order to exclude a pixel value that exceeds or does not satisfy the pixel value range specified in advance. There are limit processing to be set.

  Whether or not the pixel value is within a range that can be taken in advance is determined by whether the pixel value is expressed by a specified number of bits, or data within a pixel value range that can be expressed by a specified number of bits. It is determined on the basis of whether or not.

  The reason why the pixel value of the corrected predicted image is further corrected to fall within the pixel value range specified in advance is to select an image closer to the pattern of the input image block as the predicted image, for example. It is assumed that the noise component included in the pixel value of the predicted image is removed by the correction.

  That is, when a part of the pixel value of the predicted image deviates from the pixel value range of the input image block due to the first correction, the corrected pixel value is further corrected within the above-specified pixel value range to obtain the input. An image close to the pattern of the image block can be used as a predicted image, and as a result, a motion vector candidate that matches the motion of the input image block to be encoded can be used.

  As a specific process, the predicted image correction unit 20 counts pixels that need further correction out of the pixels of the predicted image subjected to the first correction based on a pixel value range specified in advance, and sets them in advance. The magnitude of the count value is determined based on the threshold value. At this time, the prediction image candidate data whose count value is smaller than the threshold value is further corrected. If the count value is larger than the threshold value, the prediction image candidate data is not used as the prediction image, or the prediction is performed with only the first correction. A prediction evaluation value is calculated using the image candidate data.

  Note that, as the above-described threshold, when the image feature of the predicted image corrected so as to be close to the image feature of the input image block is changed by further correction, a predicted image having an image feature closer to the input image block is searched. Therefore, the maximum reference value may be used as long as the image feature of the predicted image determined by the first correction is not changed. For example, the maximum ratio of the total number of pixels and the number of pixels that do not change the image characteristics even if the prediction image subjected to the first correction is further corrected are set.

  This threshold value may be set in advance as a fixed value, or may be a variable value that is arbitrarily set and updated from the image characteristics of the input image or the predicted image. However, in any case, a value is set in a range that does not substantially impair the image characteristics of the predicted image subjected to the first correction.

  As described above, when the predicted image corrected so that the image feature value of the predicted image in the prediction specified region matches the image feature value of the input image block exceeds the specified pixel value range, the predicted image is further corrected. With this configuration, it is possible to select an image close to the pattern of the input image by removing noise components exceeding the pixel value range specified in advance, and thus to select a prediction image with high coding efficiency.

Embodiment 2. FIG.
FIG. 6 is a block diagram showing the configuration of the image correction unit and the motion prediction unit of the video encoding apparatus according to Embodiment 2 of the present invention. The motion prediction unit 3a according to the second embodiment includes a predicted image in addition to the evaluation value calculation unit 10, the motion vector selection unit 11, the motion vector search range designation unit 12, and the predicted image region designation unit 13 described in the first embodiment. A feature value difference calculation unit (difference calculation unit) 21, an offset value setting unit 22, and an update evaluation value calculation unit 23 are included as components.

  These components 10 to 13 and 21 to 23 are configured as a program module that configures a moving image encoding program according to the present invention. As a result, the computer that executes the moving picture coding program can function as the components 10 to 13 and 21 to 23.

  In FIG. 6, the same reference numerals are given to the same components or data corresponding to or equivalent to those in FIGS. 2 and 4 shown in the first embodiment, and the other components are shown in FIG. 1. The constituent elements 4 to 9 have the same configuration, and a duplicate description thereof will be omitted.

Next, the operation will be described.
Here, processing different from that of the first embodiment will be described in detail.
The input image feature value calculation unit 18 inputs the input image data a related to the input image block to be encoded, analyzes the image feature, and calculates an input image feature value obtained by digitizing the image feature. On the other hand, the predicted image feature value calculation unit 19 inputs the predicted image data f of the predicted image designation region, analyzes the image feature, calculates a predicted image feature value j obtained by quantifying the image feature, and calculates the predicted image correction unit 20 and This is output to the predicted image feature value difference value calculation unit 21 of the motion prediction unit 3a.

  In addition, as the image feature value, the pixel average value, the pixel maximum value, the pixel minimum value, the pixel value median value, the pixel value maximum frequency value for the predicted image in the predicted image designation region and the pixel value in the input image block, The generation range of the pixel value is conceivable, but any feature value that can be adjusted to the image feature value of the input image block by the addition / subtraction operation on the prediction image in the prediction image designation region may be used.

  Further, as the image feature value, a feature value that best represents the image feature of the input image block is selected. The pixel value may be a pixel value for each of the color difference component, the RGB component, and other image signal components in addition to the luminance component.

  As a correction method, it is conceivable that the image feature value is corrected by multiplication using a weighting coefficient or the like. However, the correction of the predicted image in the predicted image designated region is performed by the addition / subtraction operation in which the AC component is not changed by the correction. Perform correction to match the image feature value.

  The predicted image correcting unit 20 inputs the input image feature value from the input image feature value calculating unit 18 in addition to the predicted image data f and the predicted image feature value j input from the predicted image feature value calculating unit 19, and the input image Based on the feature value and the predicted image feature value j, the predicted image defined by the predicted image data f is corrected. The predicted image data b of the corrected predicted image is output from the predicted image correction unit 20 to the motion prediction unit 3a.

  Specifically, it is defined by the predicted image data f so that the predicted image feature value j of the predicted image specified area specified for motion prediction within the motion vector search range matches the input image feature value of the input image data a. The predicted image in the predicted image designation area is corrected.

  In the motion prediction unit 3a, the internal evaluation value calculation unit 10 inputs the input image data a and the prediction image data b, and the prediction image feature value difference value calculation unit 21 inputs the input image data a, the prediction image data b, and the prediction image feature. The value j is input, and the motion vector search range specifying unit 12 inputs the input image data a.

  The evaluation value calculation unit 10 calculates the prediction evaluation value i of the prediction image corresponding to the input image block obtained from the input image data a in the same manner as in the first embodiment, and outputs it to the update evaluation value calculation unit 23. Specifically, using the predicted image data b, a predicted image block (with the same size as the predicted image specified region) in the predicted image specified region corresponding to the input image block to be encoded specified by the input image data a. And the difference between them is calculated as a predicted evaluation value i.

  As described above, in the present invention, since the predicted image is corrected by the addition / subtraction calculation, the corrected predicted image and the predicted image before correction have different direct current components (DC components). For this reason, the prediction image before correction corresponding to the motion vector candidate having the smallest prediction evaluation value does not always have the best encoding efficiency. In addition, there may be a plurality of motion vectors having the smallest predicted evaluation value.

  Therefore, also in the second embodiment, a threshold for evaluating the prediction evaluation value is set for the purpose of extracting some selection candidates for the prediction evaluation value in addition to the minimum prediction evaluation value. This threshold value may be set in advance as a fixed value, or may be a variable value that is arbitrarily set and updated from the image characteristics of the input image or predicted image. However, in any case, a value is set in a range in which motion vector candidates with good coding efficiency can be selected.

  For example, it is a value determined by taking statistics of parameters such as motion vector information amount, quantization step, S / N value, etc. for a plurality of different prediction evaluation values, and evaluating the image quality at the same time. Is set as the threshold value.

  The evaluation value calculation unit 10 compares the predicted evaluation value i with the threshold value every time it calculates the predicted evaluation value i, excludes the predicted evaluation value i exceeding the threshold value, extracts the predicted evaluation value i below the threshold value, and updates the evaluation value It outputs to the calculation part 23.

  The predicted image feature value difference value calculating unit 21 calculates an image feature value of the predicted image based on the predicted image data b of the corrected predicted image, and the predicted image before correction separately input from the predicted image feature value calculating unit 19. The difference value k of the predicted image feature value before and after the correction is calculated with respect to the image feature value j. The difference value k is output to the offset value setting unit 22.

  Note that as input data to the offset value setting unit 22, data obtained by classifying the difference values of the image feature values of the predicted image before and after the correction into small sections by the predicted image feature value difference value calculation unit 21 as the difference value k. In other words, any data can be used as long as the difference between the image feature values of the predicted images before and after the correction is known.

  The offset value setting unit 22 sets the offset value l based on the input difference value k between the predicted image feature values before and after correction (for example, a function value of the difference value k), and outputs it to the update evaluation value calculation unit 23. In order to select a predicted image having a pattern close to the input image block, the offset value l is set so that the closer to the image feature value of the input image block, the smaller the image feature value.

  Therefore, the offset value l tends to be smaller when the difference value k of the image feature values of the predicted image before and after correction is smaller. This is because the smaller the difference value k of the image feature values of the predicted image before and after the correction, the higher the encoding efficiency when the DC component of the predicted difference image is variable length encoded.

  Alternatively, the offset value l may be set in advance for each sequence or picture, or the offset value l may be set for each region or coding unit block divided in the picture. However, it is necessary to determine the size of the offset value l in consideration of the encoding efficiency.

  In the update evaluation value calculation unit 23, the prediction evaluation value i input from the evaluation value calculation unit 10 and the offset value l input from the offset value setting unit 22 are added, and the deviation of the prediction evaluation value before and after the correction of the prediction image is calculated. The reflected update evaluation value m is calculated and output to the motion vector selection unit 11.

  The motion vector selection unit 11 selects a motion vector that minimizes the update evaluation value m input from the update evaluation value calculation unit 23, generates motion vector information c related to the selected motion vector, and configures a subsequent configuration (image memory unit) 4 and the variable length coding unit 9).

  On the other hand, when the evaluation value calculation unit 10 extracts a plurality of prediction evaluation values i based on the threshold value, the update evaluation value calculation unit 23 calculates an update evaluation value m for each of the prediction evaluation values i. These updated evaluation values m correspond to those selected from among the predicted evaluation values i before adding the offset value l as a plurality of evaluation values, that is, those near the minimum value, from the minimum predicted evaluation value i by the above threshold. To do.

  The motion vector selection unit 11 selects a motion vector candidate corresponding to the minimum update evaluation value m among the update evaluation values m calculated for the predicted evaluation value i near the minimum value by the update evaluation value calculation unit 23.

  In this way, instead of selecting only the smallest predicted evaluation value i, a plurality of predicted evaluation values i are extracted based on the threshold value, and the predicted evaluation value i near the minimum value is obtained. It is possible to select a motion vector corresponding to a predicted image with the highest coding efficiency from among the corresponding motion vectors that can be motion vector selection candidates.

  Further, as described above, in the present embodiment, the motion vector selection unit 11 selects a motion vector based on the update evaluation value m. In this case, if there is a large difference between the offset values l for each image feature value, the possibility of selecting a motion vector corresponding to the image feature value having the smallest offset value increases.

  For this reason, if there is a large gap between the offset value of the image feature value of the predicted image close to the image feature value of the input image block and the offset value of other image feature values, motion vector candidates with better coding efficiency are discarded. There is a possibility of being.

  This will be specifically described. For example, the difference value between the pixel average value of the input image and the pixel average value of the predicted image is set as the difference value A, and the offset value obtained based on the difference value A is set as the offset value A. In addition, a difference value between the pixel value maximum frequency value of the input image and the pixel value maximum frequency value of the predicted image is set as a difference value B, and an offset value obtained based on the difference value B is set as an offset value B. Furthermore, the difference value between the pixel value median value of the input image and the pixel value median value of the predicted image is defined as a difference value C, and an offset value obtained based on the difference value C is defined as an offset value C. When the difference value magnitude relationship is difference value C <difference value B <difference value A, it is assumed that the offset value magnitude relationship is offset value C <offset value B <offset value A.

  The motion vector selection unit 11 selects a motion vector based on the update evaluation value m obtained by adding the prediction evaluation value i and the offset value l by the update evaluation value calculation unit 23. In the above example, the offset value C to be added to the predicted evaluation value i of the median pixel value closest to the input image block is smaller than the other offset values A and B. The motion vector candidate corresponding to the value is preferentially selected.

  At this time, when the offset value C is smaller in the range exceeding the allowable amount than the other offset values A and B, the update evaluation value C calculated based on the median pixel value is calculated based on the average pixel value. Compared with the update evaluation value B calculated based on the update evaluation value A and the pixel value maximum frequency value, it becomes relatively extremely small.

  This increases the probability that a motion vector corresponding to the updated evaluation value C calculated based on the pixel value median value is selected. However, if the magnitude relationship between the offset values is not considered, the predicted evaluation value that should be originally selected This includes the possibility that the motion vector corresponding to is deviated from the selection candidate and discarded.

  Therefore, the offset value l is a value that preferentially selects an image feature value of a predicted image that is close to the image feature value of the input image block, and is a value that allows a motion vector to be selected in consideration of encoding efficiency. There is a need.

  In the above-described example, in order to preferentially select the image feature value of the predicted image closest to the image feature value of the input image block, the offset value C is set to a value smaller than other offset values. At this time, in order not to be too biased toward the motion vector corresponding to the median pixel value as a motion vector candidate, the relative magnitude relationship between the offset value C and the other offset values A and B does not greatly open. Adjust to.

  That is, when the offset value setting unit 22 sets offset values corresponding to a plurality of image feature values, each offset value can take an arbitrary value (which may be expressed as an absolute value), but the relative Such a magnitude relationship is appropriately determined in consideration of the coding efficiency and adjusted to a magnitude relationship in which the motion vector having the best coding efficiency is selected.

  As described above, in the second embodiment, the difference value k between the image feature values before and after correction in the predicted image is calculated, the offset value l is set based on the difference value k, and the offset value is added to the predicted evaluation value. Since a motion vector is selected based on the update evaluation value, an image close to the design of the input image block can be selected. In addition, by encoding an image close to the design of the input image block, it is possible to select a predicted image with high encoding efficiency.

  In the second embodiment, an example in which the corrected predicted image feature value is calculated based on the predicted image data b corrected by the predicted image feature value difference value calculation unit 21 has been described. The value may be calculated outside the predicted image feature value difference value calculating unit 21 and the value of the calculation result may be input to the predicted image feature value difference value calculating unit 21. That is, in the present invention, any configuration may be used as long as the predicted image feature value difference value calculation unit 21 can calculate the difference value k of the image feature values of the predicted images before and after correction.

Embodiment 3 FIG.
FIG. 7 is a block diagram showing a motion vector selection unit and its peripheral configuration of a video encoding apparatus according to Embodiment 3 of the present invention. The motion prediction unit in the video encoding apparatus according to the third embodiment includes a priority order setting unit 24 that sets a priority order for each image feature value, instead of the offset value setting unit 22 in the second embodiment. ing.

  The priority order setting unit 24 is configured as a program module that configures a moving image encoding program according to the present invention. As a result, the computer that executes the moving picture encoding program can function as the priority order setting unit 24.

  In FIG. 7, the same reference numerals are given to the same components or data corresponding to or equivalent to those in FIG. 6 shown in the second embodiment, and the other components 2 shown in FIG. , 4 to 9 have the same configuration, and redundant description thereof will be omitted.

Next, the operation will be described.
Here, processing different from the above embodiment will be described in detail.
The priority order setting unit 24 sets a priority order for selecting a motion vector for each image feature value based on the difference value k of the predicted image feature value input from the predicted image feature value difference value calculation unit 21, Is output to the motion vector selection unit 11 as data n.

  Here, the data output by the predicted image feature value difference value calculation unit 21 to the priority order setting unit 24 may be obtained by simply obtaining the difference value k of the image feature values of the predicted image before and after correction, Data obtained by classifying the difference value of the image feature value of the predicted image before and after correction into small sections may be used. That is, any data can be used as long as the difference between the image feature values of the predicted image before and after the correction can be determined, and based on this data, the priority order for selecting a motion vector for each image feature value of the predicted image can be determined. I just need it.

  Since the priority order setting unit 24 selects a predicted image having a pattern close to the input image block, the priority order setting unit 24 sets the priority order in the order from the image feature value of the predicted image to the image feature value of the input image block. Specifically, the priority order is set to be higher with respect to the image feature value of the predicted image having a small difference value before and after correction.

  This is because a prediction image having a smaller difference value between image feature values before and after correction has higher encoding efficiency when the DC component of the prediction difference image is variable-length encoded. Note that the priority order for each image feature value of the predicted image may be set in advance for each sequence or picture, or may be set for each region divided within a picture or for each coding unit block.

  Although not shown in FIG. 7, the evaluation value calculation unit 10 includes a plurality of evaluation values based on threshold values for evaluating not only the minimum predicted evaluation value but also the predicted evaluation value, as in the first and second embodiments. The predicted evaluation value of is extracted. For example, as the threshold value, a maximum upper limit value of a prediction evaluation value that can provide an acceptable encoding efficiency is set.

  When a plurality of prediction evaluation values i are extracted below the threshold, these prediction evaluation values i are output from the evaluation value calculation unit 10 to the motion vector selection unit 11 and are not particularly described in the figure but have priority. Also output to the rank setting unit 24.

  The priority order setting unit 24 ranks the image feature values of the predicted images corresponding to the prediction evaluation values i based on the magnitude relationship of the plurality of prediction evaluation values i extracted by the evaluation value calculation unit 10 as being below the threshold value. Do. Furthermore, the priority order setting unit 24 uses the difference value k of the predicted image feature value input from the predicted image feature value difference value calculation unit 21, and corresponds to the difference value k as described above and the ranking described above. Priorities relating to motion vector selection are set for each performed image feature value.

The reason for this processing is as follows.
For example, if the predicted evaluation value i of the predicted image is larger than the threshold value even though the priority based on the difference value k is the predicted image of the first image feature value, the evaluation value calculation unit 10 The predicted evaluation value i is not output.

  In addition, in an image having a large degree of change such as luminance change or color difference change due to fade-in or flash, the difference value k of the image feature value of the predicted image before and after correction is large, and the priority order set based on this is the first priority. Even if the predicted image of the image feature value corresponding to the second or third place is the predicted evaluation value i smaller than the threshold value, it is extracted by the evaluation value calculation unit 10.

  In this case, the motion vector selection unit 11 selects a motion vector candidate corresponding to the predicted image having the second or third image feature value based on the difference value k, and the image having the first priority. The predicted image of the feature value is not selected.

  Therefore, the image feature values of the prediction images corresponding to the plurality of prediction evaluation values i extracted as the threshold value or less are ranked, and each image feature value corresponding to the difference value k and subjected to the ranking described above is ranked. Set priorities for motion vector selection.

  In the motion vector selection unit 11, if there are a plurality of motion vector candidates searched in the motion vector search range, the motion vector corresponding to the predicted image of the image feature value having the highest priority input from the priority order setting unit 24 among them. Is selected, and motion vector information c relating to the selected motion vector is generated and output to the subsequent configuration (image memory unit 4 and variable length encoding unit 9).

  As described above, according to the third embodiment, the priority order n related to motion vector selection is set for each image feature value of the predicted image in the predicted image designation region, and the predicted evaluation value calculated by the evaluation value calculation unit 10 is set. Since the motion vector is selected based on the priority order from among the motion vector candidates corresponding to, an image closer to the design of the input image block can be selected based on the priority order of the prediction evaluation value and the image feature value. In addition, by encoding an image closer to the design of the input image block, it is possible to select a predicted image with high encoding efficiency.

Embodiment 4 FIG.
FIG. 8 is a block diagram showing a motion vector selection unit and its peripheral configuration of a video encoding apparatus according to Embodiment 4 of the present invention. The motion prediction unit in the moving picture coding apparatus according to the fourth embodiment includes the priority order setting unit 24 shown in the third embodiment in the configuration of the second embodiment.

  Similar to the above-described embodiment, these components are also configured as program modules that configure the moving image encoding program according to the present invention, and the computer executes the moving image encoding program so as to function as the above-described components. be able to.

  In FIG. 8, the same reference numerals are given to the same components or data as those in FIG. 6 shown in the second embodiment, or the other components 2 shown in FIG. , 4 to 9 have the same configuration, and redundant description thereof will be omitted.

Next, the operation will be described.
Here, processing different from the above embodiment will be described in detail.
The difference value k from the predicted image feature value difference value calculation unit 21 is output to the offset value setting unit 22 and the priority order setting unit 24. The difference value k may be data obtained by classifying the difference values of the image feature values of the predicted image before and after the correction into small sections, as long as the difference between the image feature values of the predicted image before and after the correction is known. Good.

  In the offset value setting unit 22, the offset value l is set based on the difference value k between the predicted image feature values before and after correction input from the predicted image feature value difference value calculation unit 21, as in the second embodiment. (For example, the function value of the difference value k) is output to the update evaluation value calculator 23.

  Since the setting of the offset value 1 described above selects a predicted image having a pattern close to the input image block, the closer to the image feature value of the input image block, the smaller the offset value tends to be. Set to be. That is, the offset value tends to be smaller as the difference value k between the image feature values of the predicted images before and after correction is smaller.

  Further, the offset value l may be classified into small sections with respect to the image feature value, and the value may be set in a stepwise manner. Further, the offset value may be set in advance for each sequence or picture, or may be set for each region in the picture or for each coding unit block.

  The update evaluation value calculation unit 23 adds the offset value l set as described above and the predicted evaluation value i calculated by the evaluation value calculation unit 10 in the same manner as in the first embodiment to update the update evaluation value. m is calculated. The update evaluation value m is output from the update evaluation value calculation unit 23 to the motion vector selection unit 11. When the evaluation value calculation unit 10 extracts a plurality of prediction evaluation values based on the threshold value for evaluating the prediction evaluation value shown in the first embodiment, the update evaluation value calculation unit 23 has a plurality of update evaluation value calculation units 23 corresponding thereto. The update evaluation value m is calculated and output to the motion vector selection unit 11.

  The priority order setting unit 24 sets a priority order for selecting a motion vector for each image feature value based on the difference value k input from the predicted image feature value difference value calculation unit 21, and sets the priority order as data n. The result is output to the motion vector selection unit 11. Specifically, in order to select a predicted image having a pattern close to the input image block, among the image feature values of the predicted image, the image feature value of the predicted image having the smallest difference value k, in order from the image feature value of the input image block. Is set to have a higher priority.

  This is because a prediction image having a smaller difference value between image feature values before and after correction has higher encoding efficiency when the DC component of the prediction difference image is variable-length encoded. Note that the priority order for each image feature value of the predicted image may be set in advance for each sequence or picture, or may be set for each region divided within a picture or for each coding unit block.

  The motion vector selection unit 11 is set with a threshold regarding the difference between the update evaluation value m input from the update evaluation value calculation unit 23 and the minimum update evaluation value, and the update evaluation value calculated by the update evaluation value calculation unit 23 is set. Among m, the update evaluation value m whose difference from the minimum update evaluation value is equal to or less than the threshold is extracted.

  The threshold value regarding the difference from the minimum update evaluation value may be a preset value as a fixed value, or a variable value that is arbitrarily set and updated from the image characteristics of the input image or predicted image. However, in any case, a value is set within a range in which a plurality of motion vector candidates with high coding efficiency can be selected.

  For example, by changing the value of the difference between the update evaluation value m and the minimum value, taking statistics of parameters such as the amount of information of the motion vector, the quantization step, and the S / N value, and simultaneously determining the image quality, The upper limit value of the difference that gives a motion vector with good coding efficiency is set as the threshold value.

  By doing so, it is possible to select a motion vector close to the input image from among a plurality of update evaluation values m near the minimum value, and any of the update evaluation values m can derive a motion vector candidate.

  The predicted image corresponding to the update evaluation value m extracted by the threshold value may not include an image having an image feature value with the highest priority set by the difference value k. That is, the update evaluation value of the image feature value with the highest priority is not always selected based on the threshold value.

  Therefore, the motion vector selection unit 11 selects the predicted image having the highest image feature value input from the priority setting unit 24 among the image feature values of the predicted image corresponding to the update evaluation value m selected by the threshold value. A corresponding motion vector candidate is selected, and motion vector information c relating to the selected motion vector is generated and output to the subsequent configuration (image memory unit 4 and variable length coding unit 9).

As described above, according to the fourth embodiment, the offset value l is set based on the difference value k between the image feature values before and after the correction of the predicted image in the predicted image designation region (for example, a function of the difference value k). Value), an update evaluation value m obtained by adding the offset value l and the prediction evaluation value i, and a priority order for motion vector selection is set for each image feature value of the prediction image, and the update evaluation value m and the priority order are set. Since the motion vector used for encoding the input image block to be encoded is selected based on the image, the prediction evaluation value corresponding to the image feature value of the predicted image close to the image feature value of the input image block is preferentially selected. And an image close to the pattern of the input image block can be selected. Moreover, it is possible to select a prediction image with high encoding efficiency by encoding an image close to the pattern of the input image.
In the first to fourth embodiments, the apparatus of the present invention is embodied by a computer program. However, all or a part of the apparatus may be configured by hardware, which limits the present invention. is not.

It is a block diagram which shows the structure of the moving image encoder by Embodiment 1 of this invention. It is a block diagram which shows the structure of the motion estimation part in FIG. It is a figure which shows the prediction image frame memorize | stored in the image memory part in FIG. 1, the motion search range prescribed | regulated to this, a prediction image designation | designated area | region, and a prediction image block. It is a block diagram which shows the structure of the image correction part in FIG. It is a graph which shows the relationship between the predicted value of a predicted image designation | designated area | region, the pixel value of an input image, and its occurrence frequency. It is a block diagram which shows the structure of the image correction part and the motion estimation part of the moving image encoder by Embodiment 2 of this invention. It is a block diagram which shows the motion vector selection part and its periphery structure of the moving image encoder by Embodiment 3 of this invention. It is a block diagram which shows the motion vector selection part and its periphery structure of the moving image encoder by Embodiment 4 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Moving image encoder, 2 Image correction part, 3, 3a Motion estimation part, 4 Image memory part, 5 Difference calculating part, 6 Encoding part, 7 Decoding part, 8 Decoding addition part, 9 Variable length encoding part, DESCRIPTION OF SYMBOLS 10 Evaluation value calculation part, 11 Motion vector selection part, 12 Motion vector search range designation | designated part (search range designation | designated part), 13 Predictive image area designation | designated part (area | region designation | designated part), 14 Predictive image frame, 15 Motion vector search range, 16 Predicted image designation region, 17 Predicted image block, 18 Input image feature value calculating unit, 19 Predicted image feature value calculating unit, 20 Predicted image correcting unit, 21 Predicted image feature value difference calculating unit (difference calculating unit), 22 Offset value setting Part, 23 update evaluation value calculation part, 24 priority order setting part.

Claims (11)

In a video encoding device that encodes a video by motion prediction,
A search range specifying unit for specifying a search range of a motion vector;
An area designating unit for designating a predicted image designating area to be corrected for motion prediction within the motion vector search range;
As the motion prediction correction, a predicted image correction unit that corrects the image feature value of the predicted image in the predicted image designation region so as to match the image feature value of the encoding target image;
An evaluation value calculation unit that calculates a prediction evaluation value of the prediction image indicating a degree of difference between the encoding target image by a matching process between the prediction image subjected to the motion prediction correction and the encoding target image. When,
A motion image encoding device comprising: a motion vector selection unit that selects a motion vector used for encoding the encoding target image based on the prediction evaluation value. The predicted image correction unit calculates a plurality of types of image feature values in the predicted image of the predicted image designation area, calculates image feature values corresponding to the plurality of types of image features for the encoding target image, and sets these values. Correct the predicted image to fit each,
The evaluation value calculation unit calculates a prediction evaluation value of the prediction image by a matching process between the prediction image corrected for each image feature value and the encoding target image,
The moving image encoding apparatus according to claim 1, wherein the motion vector selection unit selects a motion vector based on the predicted evaluation value calculated for each image feature value by the evaluation value calculation unit.   When the prediction image correction unit performs motion prediction correction on the prediction image in the prediction image designation region, and the corrected pixel value in the prediction image exceeds an allowable pixel value range, the pixel value of the prediction image is The moving image encoding apparatus according to claim 1, further correcting so as to be within the permissible pixel value range. The evaluation value calculation unit is configured to calculate a plurality of prediction evaluation values from the smallest prediction evaluation value among the prediction evaluation values of the prediction image calculated by the matching process between the prediction image subjected to the motion prediction correction and the encoding target image. Extract the predicted evaluation value based on the threshold for extracting
When a plurality of prediction evaluation values are extracted based on the threshold value by the evaluation value calculation unit, the motion vector selection unit encodes the encoding target image from among motion vector candidates corresponding to the prediction evaluation values. 2. The moving image encoding apparatus according to claim 1, wherein a motion vector to be used is selected.   2. The moving image according to claim 1, wherein the motion vector selection unit selects a motion vector candidate whose image feature value before correction of the predicted image in the predicted image designation region is closest to the image feature value of the encoding target image. Image encoding device. A difference value calculation unit that calculates a difference value between image feature values before and after correction in the predicted image of the predicted image designation region;
An offset value setting unit that sets an offset value based on the difference value of the image feature value calculated by the difference value calculation unit;
An update evaluation value calculation unit that calculates an update evaluation value obtained by adding the offset value and the predicted evaluation value;
2. The moving image according to claim 1, wherein the motion vector selection unit selects a motion vector based on the update evaluation value from among motion vector candidates corresponding to the predicted evaluation value calculated by the evaluation value calculation unit. Image encoding device. A priority order setting unit that sets a priority order for motion vector selection for each image feature value of the predicted image in the predicted image designation region;
The motion vector selection unit selects a motion vector from motion vector candidates corresponding to the predicted evaluation value calculated by the evaluation value calculation unit based on the priority set by the priority setting unit. The moving picture coding apparatus according to claim 1. A difference value calculation unit that calculates a difference value between image feature values before and after correction in the predicted image of the predicted image designation region;
An offset value setting unit that sets an offset value based on the difference value of the image feature value calculated by the difference value calculation unit;
An update evaluation value calculation unit for calculating an update evaluation value obtained by adding the offset value and the prediction evaluation value;
A priority order setting unit that sets a priority order for motion vector selection for each image feature value of the predicted image in the predicted image designation region,
The motion vector selection unit is set by the update evaluation value calculated by the update evaluation value calculation unit and the priority order setting unit from among motion vector candidates corresponding to the predicted evaluation value calculated by the evaluation value calculation unit. 2. The moving picture encoding apparatus according to claim 1, wherein a motion vector is selected based on the priority order. When encoding a moving image by motion prediction, a search range specifying unit for specifying a search range of motion vectors,
An area designating unit for designating a predicted image designating area to be corrected for motion prediction within the motion vector search range;
As the motion prediction correction, a predicted image correction unit that corrects the image feature value of the predicted image in the predicted image designation region so as to match the image feature value of the encoding target image;
An evaluation value calculation unit that calculates a prediction evaluation value of the prediction image indicating a degree of difference between the encoding target image by a matching process between the prediction image subjected to the motion prediction correction and the encoding target image. ,
A motion vector selection unit that selects a motion vector used for encoding the encoding target image based on the prediction evaluation value from among motion vector candidates corresponding to the prediction evaluation value calculated by the evaluation value calculation unit. A program that makes it work. When encoding a moving image by motion prediction, a search range specifying unit for specifying a search range of motion vectors,
An area designating unit for designating a predicted image designating area to be corrected for motion prediction within the motion vector search range;
As the motion prediction correction, a predicted image correction unit that corrects the image feature value of the predicted image in the predicted image designation region so as to match the image feature value of the encoding target image;
An evaluation value calculation unit that calculates a prediction evaluation value of the prediction image indicating a degree of difference between the encoding target image by a matching process between the prediction image subjected to the motion prediction correction and the encoding target image. ,
A difference value calculation unit for calculating a difference value between image feature values before and after correction in the predicted image of the predicted image designation region;
An offset value setting unit that sets an offset value based on the difference value of the image feature value calculated by the difference value calculation unit;
An update evaluation value calculation unit for calculating an update evaluation value obtained by adding the offset value and the predicted evaluation value;
A computer as a motion vector selection unit that selects a motion vector used for encoding the encoding target image from the motion vector candidates corresponding to the predicted evaluation value calculated by the evaluation value calculation unit. A program that makes it work. When encoding a moving image by motion prediction, a search range specifying unit for specifying a search range of motion vectors,
An area designating unit for designating a predicted image designating area to be corrected for motion prediction within the motion vector search range;
As the motion prediction correction, a predicted image correction unit that corrects the image feature value of the predicted image in the predicted image designation region so as to match the image feature value of the encoding target image;
An evaluation value calculation unit that calculates a prediction evaluation value of the prediction image indicating a degree of difference between the encoding target image by a matching process between the prediction image subjected to the motion prediction correction and the encoding target image. ,
A difference value calculation unit for calculating a difference value between image feature values before and after correction in the predicted image of the predicted image designation region;
An offset value setting unit that sets an offset value based on the difference value of the image feature value calculated by the difference value calculating unit;
An update evaluation value calculation unit for calculating an update evaluation value obtained by adding the offset value and the predicted evaluation value;
A priority order setting unit that sets a priority order for motion vector selection for each image feature value of the predicted image in the predicted image designation region;
Based on the update evaluation value calculated by the update evaluation value calculation unit and the priority set by the priority setting unit from among motion vector candidates corresponding to the predicted evaluation value calculated by the evaluation value calculation unit. A program that causes a computer to function as a motion vector selection unit that selects a motion vector used for encoding the encoding target image.

Images