JP5299820B2 - Image processing apparatus and moving image encoding apparatus. - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To adaptively control a position of a search range window in motion vector detection when performing inter-frame prediction for an image processing apparatus. <P>SOLUTION: Motion vector detector 1 to 10 search the most similar block out of blocks included in the search range windows set in a reference image for a plurality of every blocks which are created by dividing the image of the processing object, and detect a motion vector based on the search results. Search range window setters 11 to 13 set the search range windows used by the motion vector detector following to preset rules so as to assure blocks of a predicted motion direction in the search range windows based on the motion vector detected for the block which locates in a preceding stage of a target block whose motion vector is detected. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

The present invention relates to an image processing apparatus such as an image encoding apparatus that encodes an image, and in particular, by adaptively controlling the position of a search range window in motion vector detection in inter-frame prediction, for example, motion is Image processing for reducing the amount of generated code and improving image quality by increasing the possibility that reference image data giving a smaller SAD (Sum of Absolute Difference) value is included in the search range even in motion vector detection of a large block Relates to the device.
Here, the SAD value is an absolute sum of pixel value differences.

  For example, when transmitting a video in a digital format in a video conference, digital TV broadcast, or the like, it is difficult to transmit as it is because the amount of uncompressed data is enormous. For this reason, encoding (compression) processing is usually performed before sending to the transmission line.

  Today, when encoding such moving image data, the amount of data is compressed by utilizing the high autocorrelation (redundancy) of many video signals. The redundancy of moving images includes two-dimensional spatial redundancy and temporal redundancy. Spatial redundancy refers to local similarity of pixels within a frame, and temporal redundancy refers to similarity of pixels between frames. In the moving picture coding apparatus, the data amount is compressed by reducing the redundancy.

  MPEG systems known as compression encoding / decoding technologies and H.264 are known. In the H.264 system and the like, the effect of data compression is improved by introducing an intra-frame prediction image that reduces spatial redundancy and an inter-frame prediction image that reduces temporal redundancy.

FIG. 4 shows a configuration example of a moving picture encoding apparatus having a general function.
The moving image encoding apparatus of this example includes an intra-frame prediction image generation unit 101, an inter-frame prediction image generation unit 102, a prediction image selection unit 103, and an encoding unit 104.

An example of the operation in the moving picture encoding apparatus of this example will be shown.
The image data to be encoded is input to the intra-frame predicted image generation unit 101, the inter-frame predicted image generation unit 102, and the encoding unit 104.
The intra-frame prediction image generation unit 101 generates an intra-frame prediction image, and the inter-frame prediction image generation unit 102 generates an inter-frame prediction image. From both the generated prediction images, the prediction image selection unit 103 selects a prediction image so that the generated code amount when encoding is reduced. In the encoding unit 104, the difference between the selected predicted image and the input image data that is the prediction target image is obtained, and variable length encoding processing is performed. Thereby, encoded data is generated.

FIG. 5 shows a configuration example of the inter-frame prediction image generation unit 102 having a conventional motion vector detection function. In this example, the case where the inter-frame prediction image generation unit 102 includes the encoding unit 119 having the same function as the encoding unit 104 illustrated in FIG. 4 is illustrated.
The inter-frame predicted image generation unit 102 of the present example includes an input image data holding memory 111 that is held in units of frames, a prediction target block dividing unit 112, a prediction target block data holding memory 113 that is held in units of MB (macroblocks), Reference image data holding memory 114, line memory (cache) 115 held in MBL (macroblock layer) units, search range image data holding memory 116 held in units of several tens of MB, motion vector detection unit 117, motion compensation unit 118 , An encoding unit 119 and a local decoding unit 120 are provided.

An example of the operation in the inter-frame prediction image generation unit 102 in the moving image encoding apparatus of this example will be shown.
The operation of this example is schematically shown in FIG. 1 except that the operation is improved by the processing units 11 to 13 added in the configuration shown in FIG. It is the same as the operation of the configuration.

The local decoding unit 120 locally decodes the encoded result (encoded data), and stores the locally decoded image data in the reference image data holding memory 114 for use as reference image data.
In the motion vector detection unit 117, as shown in FIG. 6, reference image data that is most similar to a block (macroblock in this example) 131 on a certain input image data (that is, the SAD value is minimized). The upper block (macroblock in this example) 132 is detected, and a vector representing the movement direction and movement amount of the reference image data block relative to the input image data block is detected as the motion vector 133.
FIG. 6 shows an example of a motion vector.

In motion vector detection in applications that require low delay, the search range of motion vectors is limited due to processing time limitations.
In such a case, in the conventional method, for example, as shown in FIGS. 7A, 7B, and 7C, prediction target blocks 142 and 152 having M × N pixels (M and N are integer values, respectively). , 162 are set as motion vector search range windows in the fixed regions 143, 153, and 163 in the vicinity of 162, and motion vectors are detected.

  FIGS. 7A, 7 </ b> B, and 7 </ b> C show examples of positional relationships between the M × N pixel prediction target blocks 142, 152, and 162 and the motion vector search range windows 143, 153, and 163, respectively. In addition, examples of holding areas 141, 151, 161 of the line memory 115 are shown.

JP 2007-243366 A Japanese Patent No. 3020299

  However, in the above-described conventional motion vector detection, the positional relationship between the M × N pixel prediction target block and the motion vector search range window is a fixed region with respect to the M × N pixel prediction target block. As shown in FIGS. 8 (a) and (b), when the prediction target blocks 171 and 181 having M × N pixels move greatly between frames, a region that gives a smaller SAD value (in this example, a macro Blocks 172 and 182 are more likely to be outside the search range. Thereby, the SAD value in the inter-frame prediction of the prediction target block of M × N pixels increases, and the prediction accuracy decreases. As a result, the amount of generated codes increases and image quality deteriorates.

FIGS. 8A and 8B show examples in which the region that gives the minimum SAD value is outside the motion vector search range.
FIG. 8A shows an example in which the prediction target block 171 having M × N pixels greatly moves upward, and FIG. 8B shows a prediction target block 181 having M × N pixels. An example is shown in which moves greatly downward.

  The present invention has been made in view of such a conventional situation, and by adaptively controlling the position of the search range window in motion vector detection at the time of inter-frame prediction, for example, the motion of a block having a large motion An object of the present invention is to provide an image processing apparatus capable of reducing the amount of generated codes and improving image quality by increasing the possibility that reference image data giving a smaller SAD value is included in the search range even in vector detection. And

In order to achieve the above object, in the present invention, the image processing apparatus has the following configuration.
That is, the motion vector detection means searches for the most similar block from among the blocks included in the search range window set in the reference image for each of a plurality of blocks generated by dividing the processing target image. Thus, a motion vector is detected based on the search result.
The search range window setting means performs prediction based on the motion vector detected by the motion vector detection means for a block located in the preceding stage with respect to a block whose motion vector is detected by the motion vector detection means. The search range window used by the motion vector detecting means is set in accordance with a pre-set rule so as to secure a block in the search direction within the search range window.

  Therefore, when a motion vector is detected for a certain block, the search range window is set so as to secure a block in the predicted motion direction within the search range window based on the motion vector detected for the block located in the preceding stage. In this way, by adaptively controlling the position of the search range window in motion vector detection during inter-frame prediction, for example, even in motion vector detection of a block with large motion, it is smaller The possibility that the reference image data giving the SAD value is included in the search range can be increased, and the amount of generated codes can be reduced and the image quality can be improved.

Here, various image processing apparatuses may be applied.
Various images to be processed may be used. For example, an image for each frame of a moving image can be used.
Various images may be used as the reference image. For example, an image of a past frame may be used as compared with the image to be processed.
Various search range windows may be used. For example, the search range window includes a block for which a motion vector is detected, and a predetermined number (having a fraction) in each of the vertical and horizontal directions when viewed from the block. In other words, a rectangular area composed of a plurality of (may have fractional) blocks as a whole can be used (see, for example, FIG. 3).

Moreover, as a block, a macroblock can be used, for example.
Further, as a mode of detecting a motion vector for each of a plurality of blocks, for example, a mode of detecting a motion vector for all blocks may be used, or motion vectors may be used for some blocks (only). An aspect of detecting may be used.
In addition, as a mode of searching for a block most similar to a certain block, for example, a SAD value is calculated for each of the certain block and another block, and the other block having the smallest SAD value is most similar to the block. A mode of detecting as can be used.

In addition, any block may be used as the block positioned before the block that is the target of motion vector detection. For example, a predetermined number of blocks in the previous stage, such as one line in the previous stage, is used. (See, for example, FIG. 3).
Moreover, as a motion vector detected about the block located in the front | former stage from the block used as the object which detects a motion vector, the average value about a some block may be used, for example.
In addition, for example, a component (only) in a predetermined direction may be used as a motion vector detected for a block located in a stage prior to a block that is a target for detecting a motion vector or a predicted motion direction. As an example, a component (only) in the update direction of the search range window can be used (for example, see FIG. 3).

In addition, various rules may be used as the pre-set rule to ensure the block of the predicted motion direction in the search range window. For example, the pre-stage of the block that is the target of detecting the motion vector. It is possible to use a rule that determines how to set the search range window based on the magnitude relationship between a motion vector (which may be an average value) detected for a block located at and one or more predetermined threshold values. (For example, see FIG. 3).
Note that this rule is set in advance (thinking) so that a block in the predicted motion direction is reserved (rather than doing nothing) within the search range window. It may not be realized.

  In addition, when a search range window is set according to a pre-set rule so as to secure a block of the predicted motion direction in the search range window, and a target for detecting a motion vector without using such a rule A configuration in which a predetermined fixed area is set as a search range window with respect to the position of the block to be switched may be used. As an example, for an image of one frame, the first first predetermined number of blocks and the last second predetermined number of blocks (which may be the same as or different from the first predetermined number) Instead of using the rule, a configuration in which the rule is used in a block between them can be used (see, for example, step S3 and step S7 in FIG. 2).

  As described above, the image processing apparatus according to the present invention adaptively controls the position of the search range window in motion vector detection at the time of inter-frame prediction, for example, motion vector detection of a block with large motion. In this case, it is possible to increase the possibility that reference image data giving a smaller SAD value is included in the search range, and to reduce the amount of generated codes and improve the image quality.

It is a figure which shows the structural example of the estimated image generation part between frames which concerns on one Example of this invention. It is a figure which shows an example of the procedure of the motion vector search process performed in the inter-frame estimated image generation part. (A)-(d) is a figure for demonstrating the example of the relationship between the prediction object block of MxN pixel, and a line memory holding area. It is a figure which shows the structural example of a moving image encoder. It is a figure which shows the structural example of the estimated image generation part between frames. It is a figure which shows an example of a motion vector. (A), (b), (c) is a figure which shows the example of the positional relationship of the prediction object block of MxN pixel, and a motion vector search range window. (A), (b) is a figure which shows an example in case the area | region which gives the minimum SAD value becomes out of the search range of a motion vector.

Embodiments according to the present invention will be described with reference to the drawings.
FIG. 1 shows a configuration example of an inter-frame prediction image generation unit according to an embodiment of the present invention.
Here, the inter-frame prediction image generation unit of the present example is applied to, for example, the inter-frame prediction image generation unit 102 of the moving picture encoding apparatus as shown in FIG. In the present example, the case where the encoding unit 9 having the same function as the encoding unit 104 shown in FIG. 4 is provided in the inter-frame prediction image generation unit is shown.
In this example, the moving picture coding apparatus is H.264. Although the case where encoding is performed using the H.264 encoding method will be described, the encoding method is not limited to this example, and various types may be used.

  The inter-frame predicted image generation unit of this example includes an input image data holding memory 1 that holds each frame, a prediction target block dividing unit 2, a prediction target block data holding memory 3 that holds MB units, and a reference image data holding memory 4, a line memory (cache) 5 held in MBL units, a search range image data holding memory 6 held in units of several tens of MBs, a motion vector detection unit 7, a motion compensation unit 8, an encoding unit 9, and a local decoding unit 10, a line motion vector holding memory 11, a line average motion vector calculation unit 12, and a reference image data management unit 13 are provided.

An example of operation | movement in the inter-frame estimated image generation part of this example is shown.
Image data input from the outside is input to the input image data holding memory 1 and the encoding unit 9.
In the input image data holding memory 1, the inputted input image data is written and held (stored) in units of frames.
The prediction target block dividing unit 2 inputs the input image data written in the input image data holding memory 1 and cuts the input image data vertically and horizontally into M × N pixel prediction target blocks (in MB units). To divide.
The prediction target block data holding memory 3 stores and holds the data of the divided M × N pixel prediction target blocks.

Simultaneously (in parallel) with the above processing, reference image data for several slices is written to the line memory 5 from the reference image data holding memory 4 for holding reference image data, and further, reference images for several tens of MB are included therein. The data is written and held in the search range image data holding memory 6 as search range image data (motion vector search range window data).
Here, the line memory 5 is located between the reference image data holding memory 4 and the search range image data holding memory 6 and has a role of a cache memory.

  The motion vector detection unit 7 reads M × N pixel prediction target block data from the prediction target block data holding memory 3 and also reads search range image data from the motion search range image data holding memory 6 to predict M × N pixels. A process of detecting a motion vector from the target block data and the search range image data is executed. Specifically, the motion vector detection unit 7 calculates an SAD value between the prediction target block data of M × N pixels and the search range image data (reference image data) for each divided block unit. A minimum value is detected from the calculated SAD value, and a vector at the detected minimum value is detected as a motion vector (see, for example, FIG. 6).

The motion vector detected by the motion vector detection unit 7 is input to the motion compensation unit 8 and the line motion vector holding memory 11.
The motion compensation unit 8 identifies a compensation block (motion compensation block) based on the motion vector in accordance with the motion vector input from the motion vector detection unit 7, and the data of the identified compensation block is stored in a search range image data holding memory. 6, inter-frame prediction image data is generated based on the compensation block, and is output to the encoding unit 9.
Here, regarding the compensation block, a motion vector of a prediction target block of M × N pixels is searched in the search range image data, and a block specified by the obtained motion vector is set as a compensation block.

The encoding unit 9 generates the difference image data from the generated predicted image data and the input image data (which corresponds to the predicted image data) that is the basis for generating the predicted image data, and Then, the generated difference image data is encoded to generate encoded data.
The generated encoded data is output to the outside and sent to the local decoding unit 10 and input.

Local decoding unit 10 for performing local decoding is decrypt the input encoded data (encoded image data). The image data obtained in this way is written and held in the reference image data holding memory 4 as reference image data.
The data written in the reference image data holding memory 4 is transferred to the line memory 5 as necessary.

The line motion vector holding memory 11 holds a motion vector for one slice based on the motion vector input from the motion vector detection unit 7.
The line average motion vector calculation unit 12 calculates the average of motion vectors for one slice (average motion vector) based on the motion vector for one slice held in the line motion vector holding memory 11, and manages reference image data To the unit 13.

The reference image data management unit 13 controls communication between the reference image data holding memory 4 and the line memory 5 based on the average motion vector input from the line average motion vector calculation unit 12.
In this example, the reference image data management unit 13 predicts a motion vector of an M × N pixel prediction target block based on the average motion vector, and sets a motion vector set in the reference image data based on the prediction result. Adaptively controls the search range window. Thereby, the reference image data to be held is adaptively controlled with respect to the prediction target block of M × N pixels.

FIG. 2 shows an example of a flow of motion vector search processing performed in the inter-frame predicted image generation unit of this example.
First, input image data is taken into the input image data holding memory 1 (step S1).
Next, the prediction target block dividing unit 2 cuts out the data written in the input image data holding memory 1 in MB units from the head, and predicts the data of the prediction target block of M × N pixels obtained thereby. Write to the target block data holding memory 3 (step S2).

Next, it is evaluated whether or not [{number of vertical blocks of search range window (SCB) / 2}> slice number of prediction target block] (step S3).
As a result of the evaluation in step S3, if it is determined to be true, a fixed region search range window (motion vector search range window) is set for the prediction target block (step S4), and the set search is performed. A motion vector search is executed within the range (step S5).

On the other hand, if it is determined that the result of the evaluation in step S3 is false, [slice number of prediction target block <{number of vertical blocks of frame (FCB) −number of vertical blocks of search range window (SCB) / 2}}. ] Is evaluated (step S7).
As a result of the evaluation in step S7, if it is determined to be true, a search range window is adaptively set (step S8), and a motion vector search is executed within the set search range (step S9).

  On the other hand, if it is determined that the result of the evaluation in step S7 is false, a fixed area search range window is set for the prediction target block (step S10), and a motion vector search is performed within the set search range. Is executed (step S11).

After the motion vector search is executed in any of step S5, step S9, and step S11, it is checked (inspected) whether the prediction target block to be processed is the final block of the frame (step S6).
If the result of the check in step S6 is true, the motion vector search process is terminated.
On the other hand, if it is determined that the result of the check in step S6 is false, the process proceeds to the process for the next prediction target block (returns to the process in step S2).

Here, in this example, the reference image data management unit 13 performs processing such as setting of a motion vector search range window as in the processing of step S3, step S4, step S7, step S8, step S10, and step S6. Done.
In this example, the motion vector detection unit 7 performs a process for searching for (detecting) a motion vector as in the processes of steps S5, S9, and S11.

Next, a specific example of a method for adaptively setting a motion vector search range window for the flow shown in FIG.
FIG. 3 shows an example of the relationship between the prediction target block of M × N pixels and the line memory holding area.
In this example, a case where the number of vertical blocks in one frame is 34 and the number of vertical blocks in the search range window is 5 will be described as an example. In this example, it is assumed that slice numbers are assigned as 0, 1, 2,.

In the process of step S3 shown in FIG. 2, [{number of vertical blocks in search range window (SCB) / 2}> slice number of prediction target block] is false because the slice number of the prediction target block is 3 ( This is when the fourth slice) is reached. In addition, in the process of step S7 shown in FIG. 2, a prediction in which [slice number of prediction target block <{number of vertical blocks of frame (FCB) −number of vertical blocks of search range window (SCB) / 2}] ”is true. Among the slice numbers of the target block, the maximum value is 31 (the 32nd slice).
Therefore, in this example, the adaptive search range window setting process is executed when the slice number of the prediction target block is between 3 and 31.

In this example, the motion vector of each prediction target block is sequentially written in the line motion vector holding memory 11 in units of slices from the start of the motion vector search process.
As shown in FIG. 3A, the motion vector of each prediction target block of the second slice is held in the line motion vector holding memory 11 when searching for the motion vector of the third slice.
FIG. 3A shows a holding area 21 of the line memory 5, a prediction target block 22, a motion vector search range window 23, and a motion vector 24 for 1 MBL.

Since there is a correlation between the motion vectors of the blocks between adjacent slices, the tendency of the motion vector of the prediction target block belonging to the next slice is predicted using this.
In this example, the line average motion vector calculation section 12 is, according to (Equation 1), preceding slice # (In our example, 2 slices th) averaged Y direction average Y direction component Y _n of motion vectors The value (average motion vector) Y _ave is calculated. In (Expression 1), m is the total number of blocks in one slice (in this example, 1 MBL).

Here, as shown in FIG. 3, in this example, the horizontal direction in FIG. 3 is the X direction, and the vertical direction is the Y direction. In addition, the Y direction is parallel to the direction in which the plurality of MBLs are arranged, and the opposite direction to the direction along slice numbers 0, 1, 2,... Is positive. That is, it is predicted that the Y direction average value Y _ave increases in the positive direction in FIG. 3 (in the direction in which the slice number decreases) as the Y direction average value Y _ave increases, and the lower in the Y direction average value Y _{ave in the} direction in FIG. It is predicted to move in the direction of increasing).

This Y-direction average value Y _ave indicates the tendency of the Y-direction motion vector of the previous slice (second slice in this example), and based on this, the slice of the next stage (example here) Then, the tendency of the Y-direction motion vector of the third slice) is estimated.

In this example, the number of vertical blocks in the search range window is 5, which is 80 pixels (in this example, M × N = 16 × 16).
On the other hand, a threshold value (for example, −16, 16) is set, and the reference image data management unit 13 compares the threshold value with the Y-direction average value Y _ave . Based on the comparison result, the reference image data management unit 13 controls the update frequency of the data held in the line memory 5 to shift the motion vector search range window update amount (in the direction of increasing the slice number). Amount) is adaptively controlled.

In this example, the following controls (Control 1) to (Control 3) are performed. Here, it is assumed that the Y-direction average value Y _ave is expressed using the unit of the number of pixels.
(Control 1) When the Y-direction average value Y _ave is larger than 16, the data held in the line memory 5 is not updated as shown in FIG.
(Control 2) When the Y-direction average value Y _ave is not less than −16 and not more than 16, the data held in the line memory 5 is updated by one slice as shown in FIG.
(Control 3) When the Y-direction average value Y _ave is smaller than −16, the data held in the line memory 5 is updated by two slices as shown in FIG.

3 (b), (c), and (d) show line memory holding areas 31, 41, 51, prediction target blocks 32, 42, 52, and motion vector search range windows 33, 43, 53, respectively. It is. Here, in this example, the holding areas 31, 41, 51 of the line memory 5 are for 6 slices, and in the Y direction, 5 slices from the head are set as the motion vector search range windows 33, 43, 53. In the X direction, a predetermined number (may have a fraction) of blocks is set as the motion vector search range windows 33, 43, 53.
Similarly, for the subsequent slices (in the present example, the fourth and subsequent slices), the Y-direction average value Y _ave is obtained, and the data held in the line memory 5 is controlled based on the comparison result with the threshold value.

As described above, in this example, the position of the search range window in the motion vector detection at the time of inter-frame prediction is adaptively controlled, so that, for example, even in the motion vector detection of a block with a large motion Therefore, it is possible to improve the possibility that the reference image data giving a smaller SAD value is included in the search range, to improve the SAD value, and to reduce the amount of generated codes and improve the image quality.
This example is a technique suitable for encoding image data such as moving images, and is applied to various devices such as a moving image encoding device, a motion vector detection device, an inter-frame encoding device, and the like. Is possible.

  Note that the inter-frame prediction image generation unit (an example of an image processing apparatus) of this example uses the functions of the processing units 1 to 10 to perform a process of detecting a motion vector using a search range window set in a reference image. The vector detection means is configured, and the functions of the processing units 11 to 13 perform processing for setting a search range window based on a motion vector detection result for a block located before the block that is a target for detecting a motion vector. Search range window setting means is configured.

Here, the configuration of the system and apparatus according to the present invention is not necessarily limited to the configuration described above, and various configurations may be used. The present invention can also be provided as, for example, a method or method for executing the processing according to the present invention, a program for realizing such a method or method, or a recording medium for recording the program. It is also possible to provide various systems and devices.
The application field of the present invention is not necessarily limited to the above-described fields, and the present invention can be applied to various fields.
In addition, as various processes performed in the system and apparatus according to the present invention, for example, the processor executes a control program stored in a ROM (Read Only Memory) in hardware resources including a processor and a memory. A controlled configuration may be used, and for example, each functional unit for executing the processing may be configured as an independent hardware circuit.
The present invention can also be understood as a computer-readable recording medium such as a floppy (registered trademark) disk or a CD (Compact Disc) -ROM storing the control program, and the program (itself). The processing according to the present invention can be performed by inputting the program from the recording medium to the computer and causing the processor to execute the program.

1, 111 .. Input image data holding memory, 2, 112 .. Prediction target block dividing unit, 3, 113 .. Prediction target block data holding memory, 4, 114 .. Reference image data holding memory, 5, 115. Line memory (cache), 6, 116 .. Search range image data holding memory, 7, 117... Motion vector detection unit, 8, 118... Motion compensation unit, 9, 104, 119. 120 .... Local decoding unit, 11 .... Line motion vector holding memory, 12 .... Line average motion vector calculation unit, 13 .... Reference image data management unit,
21, 31, 41, 51, 141, 151, 161... Holding area of line memory 22, 32, 42, 52, 142, 152, 162, 171, 181... Prediction target block, 23, 33, 43, 53, 143, 153, 163 .. motion vector search range window, 24 .. 1 MBL worth of motion vectors,
101 .. Prediction image generation unit within frame, 102.. Prediction image generation unit between frames, 103.. Prediction image selection unit,
131, 132 ... blocks, 133 ... motion vectors,
172, 182... Region giving a small SAD value,

Claims (3)

For each of the plurality of blocks generated by dividing the image to be processed, the most similar region is searched from the search range window set in the reference image, and the motion vector is detected based on the search result. Motion vector detecting means for performing,
A block having a motion direction predicted based on a motion vector detected by the motion vector detection unit for a block to be processed before a block for which a motion vector is detected by the motion vector detection unit. Search range window setting means for setting the search range window used by the motion vector detection means according to a pre-set rule to ensure within the search range window;
In an image processing apparatus comprising:
One macroblock line, which is a horizontal arrangement of macroblocks, is defined as one slice,
A reference image data holding memory for holding a reference image in units of frames;
A line memory that caches reference image data for a plurality of slices transferred from the reference image data holding memory;
A search range image data holding memory for holding reference image data for a plurality of macroblocks constituting the search range window from a line memory,
The motion vector detection means detects the motion vector from reference image data read from the search range image data holding memory,
The search range window setting means determines the predicted motion in the next slice of the one slice based on a plurality of motion vectors detected in one slice by the motion vector detection means, and the line memory An image processing apparatus characterized by adaptively controlling a slice to be cached . A moving image encoding apparatus that includes the image processing apparatus according to claim 1 and performs interframe predictive encoding,
  An input image data holding memory for holding the image to be processed;
  A prediction target block data holding memory that holds, in macroblock units, a prediction target macroblock cut out from the processing target image held by the input image data holding memory;
  The process of writing from the reference image data holding memory to the line memory is performed in parallel with the process of writing from the input image data holding memory to the prediction target block data holding memory,
    The search range window setting means determines the average of the vertical components of a plurality of motion vectors detected in the one slice as the predicted motion in the next slice, and sets the center in the vertical direction of the search range window. A moving picture coding apparatus that performs control to match a positional relationship between a position and a macroblock to be predicted to the predicted motion. The moving image encoding apparatus according to claim 2, wherein the search range window is a horizontally long rectangle.

Images