CN1784016A - Intra prediction apparatus - Google Patents

Abstract

The present invention provides an intra-prediction device that helps to improve picture quality and coding efficiency, and can reduce calculation load. The intra prediction means (11) includes a candidate limiting unit (111) for defining candidates of intra prediction directions applicable to a plurality of pixel blocks constituting the image data from among a plurality of intra prediction directions by adapting to a feature of the image data, and using An intra prediction execution unit (112) that performs intra prediction based on the intra prediction direction defined by the candidate definition unit (111). Thereby, while contributing to improvement of image quality and coding efficiency, calculation load can be reduced.

Description

internal predictor

technical field

The present invention relates to an internal prediction device that can be used for animation compression, etc., and also relates to a technology for reducing the calculation amount of internal prediction as an animation compression method.

Background technique

In recent years, with the popularization of communication infrastructure such as broadband and the decline in the prices of large-capacity storage devices such as personal computers, HDDs, and DVDs, memory cards, etc., the environment where ordinary consumers edit, store, transmit, and carry images Already available, more and more users.

In this way, on the one hand, the scenes of image processing by ordinary consumers are increasing, but on the other hand, even if the performance of personal computers is continuously improved, better performance in image processing cannot be obtained. For example, if the amount of image information is very large, it takes a lot of time to compress a movie even with a high-performance personal computer. As a strategy to solve this problem, high-efficiency computation methods that can handle the same encoding performance with less computation can be considered. In addition, if high-efficiency calculations are performed, the restriction of the battery life of portable photographic equipment on the photographing time can be alleviated due to the reduction of power consumption. Based on this background, more efficient animation compression techniques can be continuously sought.

Among them, the newly-produced international standard animation compression standard H.264 (for example, refer to Non-Patent Document 1) aims at improving image quality and coding efficiency, and adopts various animation compression methods. In addition, H.264 has the feature of sequentially comparing multiple prediction methods and selecting the prediction method with the best coding efficiency for encoding.

For example, in the case of encoding a 16×16-pixel macroblock included in picture I shown in FIG. , perform intra-prediction on the macroblock from multiple directions.

As a typical example, 4×4 intra prediction compares multiple prediction methods and selects the prediction method with the best coding efficiency.

In this prediction method, as shown in FIG. 2, there are prediction mode 0 (vertical) which calculates the predicted pixel value from the upper MB (macroblock) and predicts the pixel value in the vertical direction, and calculates the predicted pixel value from the adjacent MB and predicts the horizontal direction. Prediction mode 1 (horizontal) for directional pixel values, prediction mode 8 (horizontal-up), prediction mode 6 (horizontal- Bottom), Prediction Mode 4 (Diagonal-Bottom-Right), Prediction Mode 5 (Vertical-Right), Prediction Mode 7 in which predictions are made in directions ±22.5-44.5 degrees off vertical by calculating predicted pixel values from adjacent MBs (Vertical-left), prediction processing in 8 directions of prediction mode 3 (diagonal-bottom-left), plus prediction mode 2 (average) which performs prediction based on the average of adjacent MB pixel values, a total of 9 types.

Likewise, 16×16 intra prediction also has multiple prediction methods. There are four prediction methods of the same type in this 16×16 intra prediction.

For 16×16 intra prediction, as shown in FIG. 3 , there are prediction methods of prediction mode 0 (vertical), prediction mode 1 (horizontal), prediction mode 2 (average), and prediction mode 3 (planar).

Therefore, in the conventional intra prediction apparatus, as shown in FIG. 4 , when picture I is input, prediction errors (absolute value difference sums) related to nine prediction modes are calculated for block 0 . That is, for all nine kinds of prediction modes, the difference value and the sum of absolute value differences between the reference pixel and the pixel to be encoded are calculated. Thus, the intra prediction means uses the prediction with the smallest prediction error as the internal prediction.

Once the intra prediction of block 0 is finished, the intra prediction means returns to the same processing as described above, and performs intra prediction of blocks 1 to 15 .

Once the intra prediction of the block 15 is completed, the difference value and the absolute value difference sum between the reference pixel and the pixel to be encoded are calculated in all four prediction modes for the macroblock. Thus, the intra prediction means uses the smallest prediction error, that is, uses, for example, the smaller one of the sum of absolute value difference sums at respective blocks and the absolute value difference sum of macroblocks as intra prediction. That is, in the case where the total sum of absolute value differences for each block is small, the intra prediction means outputs the difference value for each block. On the contrary, when the absolute value difference sum of the macroblocks is small, the intra prediction device outputs the difference value of the relevant macroblocks.

In this way, data compression is performed on the picture I.

However, as shown in the above example, in the conventional intra prediction device, in the intra prediction of H.264, which is the international standard currently disclosed, the video compression method H.264 adopts various methods for improving picture quality. Prediction methods (9 types in 4×4 intra prediction, 4 types in 16×16 intra prediction) are used to create a predicted image. Since the most suitable prediction method is selected from various prediction methods, that is, due to the result of encoding The process with the highest encoding efficiency is selected, so the increase in the amount of encoding calculations caused by the improvement of image quality and the improvement of encoding efficiency cannot be avoided, thereby increasing the overhead and calculation load.

Therefore, in order to realize high-quality and high-efficiency video compression technology, it is necessary to reduce the amount of H.264 encoding calculations that optimize image quality.

[Non-patent literature] "H.264 Advanced Video Coding for Generic Audiovisual Services" ("H.264Advanced video coding for generic audiovisual services", published by トリケツプスPublishing House).

Contents of the invention

Here, an object of the present invention is to provide an intra prediction device capable of reducing calculation load while contributing to improvement of image quality and coding efficiency.

In order to achieve the related object, the present inventors have found after research that: in the case of usually using pixels of adjacent macroblocks (abbreviated as MB hereinafter) for prediction, the correlation between pixels becomes smaller than the distance between pixels of adjacent MBs. get lower. In the case of intra-prediction using all prediction directions regardless of inter-pixel correlation, the present inventors noticed that intra-prediction using prediction directions that do not contribute to coding efficiency would result in useless processing. Also, the present inventors have noticed that if intra prediction is performed without checking image characteristics, unnecessary processing such as prediction direction processing that does not contribute to coding efficiency is performed.

In addition, there are the following problems in the method of systematically handling prediction directions to reduce internal prediction processing. For example, in 4×4 intra prediction, the prediction in the horizontal direction and the vertical direction is first evaluated, and based on the result, the prediction direction for the next evaluation is determined. In this scheme, the inventors noticed that even if it is known in advance that the prediction in the vertical direction does not contribute to the coding efficiency, re-evaluation in the vertical direction is required, so there is a problem of performing prediction in the unnecessary vertical direction.

Based on these points of attention, the present invention was conceived.

In order to solve the above-mentioned problems, the intra prediction device according to the present invention is characterized in that it includes a method for limiting candidate intra prediction directions suitable for a plurality of pixel blocks constituting the image data from among a plurality of intra prediction directions adapted to the characteristics of the image data. candidate defining means, and intra prediction performing means for performing intra prediction using the intra prediction direction defined by said candidate defining means.

That is, the present invention omits in advance processes that are harmful to encoding efficiency and processes that do not make a difference in encoding efficiency according to the characteristics of images when performing intra prediction.

Thus, while contributing to the improvement of image quality and coding efficiency, the calculation load can be reduced.

In the intra prediction device according to the present invention, the candidate limiting means can limit candidates for an intra prediction direction based on a pixel aspect ratio that is a feature of the image.

In addition, in the intra prediction device related to the present invention, the aspect ratio characteristic can be obtained from any one of image signal input equipment and information obtained from inside and outside of the video stream.

Also, in the intra prediction device according to the present invention, the candidate limiting means is characterized by using the horizontal and vertical distances between adjacent pixels in the original image according to the pixel aspect ratio as the horizontal and vertical distances of the intra prediction. Confidence of predictions to limit candidates for internal prediction directions.

In addition, in the intra prediction device according to the present invention, the candidate limiting means is characterized in that the intra prediction can be limited based on picture structure information characteristic of the image data by indicating whether it is either interlaced scanning or progressive scanning. alternate direction.

In addition, in the intra prediction device according to the present invention, the picture structure information is characterized in that it can be obtained from any one of the image signal input device and information obtained from inside or outside the video stream.

In addition, in the intra prediction device according to the present invention, the candidate limiting means is characterized in that, when the picture structure information indicates interlaced scanning, the vertical direction included in the intra prediction direction can be preferentially excluded from the candidates. -right direction and vertically-left direction.

In addition, in the intra prediction device according to the present invention, the candidate limiting means is characterized in that, when the picture structure information indicates interlaced scanning, the vertical direction included in the intra prediction direction can be preferentially excluded from the candidates. - right direction, vertical-left direction, and diagonal-bottom-right direction and diagonal-bottom-left direction.

In addition, in the intra prediction device according to the present invention, the candidate limiting means is characterized in that, when the picture structure information indicates interlaced scanning, an average value included in the intra prediction direction can be preferentially excluded from the candidates. direction.

In addition, in the intra prediction device according to the present invention, the candidate limiting means can preferentially exclude an average direction included in the intra prediction direction from the candidates when the picture structure information indicates interlaced scanning. and vertical direction.

In addition, in the intra prediction device according to the present invention, the candidate limiting means is characterized in that when encoding is performed using a picture-level adaptive (Pict-Level Adaptive) related to a moving picture compression method, it is possible to limit Candidates for the internal forecast direction.

In addition, in the intra prediction device according to the present invention, the picture structure information is any one of field structure and frame structure, and the candidate limiting means is characterized in that it uses macroblock-level adaptive ( MB-Level Adaptive) encoding, when predicting by switching a pair of macroblock pairs on the field structure and a pair of macroblock pairs on the frame structure, it can correspond to the information of whether the macroblock pair is a field structure or a frame structure Limit internal forecast candidates.

In addition, in the intra prediction device according to the present invention, the candidate limiting means can limit candidates for an intra prediction direction based on a result obtained for an original picture or a coded picture having a preceding encoding order.

According to the above, candidates for intra prediction directions can be easily limited.

Also, in the intra prediction device according to the present invention, it is possible to use an original picture or a coded picture immediately preceding a picture to be coded as an original picture or a coded picture preceding in coding order.

As a result, the correlation between the preceding original image or encoded image and the image to be encoded can be increased, and the intra prediction direction can be precisely defined.

In addition, in the intra prediction device according to the present invention, the result is a motion vector obtained from a coded picture preceding in coding order, and the candidate limiting means can limit the direction of the intra prediction based on the direction of the motion vector. alternate.

As a result, when a moving image exists in an image, it is possible to accurately and efficiently exclude intra-prediction direction candidates with low correlation from known motion vectors.

In addition, in the intra prediction device according to the present invention, it is characterized in that, as the motion vector, any data representing the spatial same position as the macroblock to be coded, positions including the surroundings of the spatial same position, and the entire image can be used. A motion vector for a unit area.

As a result, the accuracy of the motion vector for the macroblock to be coded can be improved, and intra prediction direction candidates with low correlation can be accurately and efficiently excluded from the motion vector.

In addition, in the intra prediction device according to the present invention, the result is an intra prediction direction obtained from a coded picture preceding in coding order, and the candidate limiting means can limit the intra prediction direction based on the intra prediction direction. alternate.

As a result, it is possible to accurately and efficiently exclude intra prediction direction candidates with low correlation from known intra prediction directions without being restricted by image motion conditions.

In addition, in the intra prediction device of the present invention, it is characterized in that, as the intra prediction direction, an intra prediction direction related to a macroblock at the same spatial position as the current macroblock and a position including the surroundings of the spatial same position can be used. .

As a result, the accuracy of the intra prediction direction for the current macroblock can be improved, and intra prediction direction candidates with low correlation can be accurately and efficiently excluded from the intra prediction direction.

In addition, in the intra prediction device according to the present invention, the result is an adjacent pixel difference of the original image or encoded image preceding the encoding order, and the candidate limiting means can limit the intra prediction based on the adjacent pixel difference. direction alternate.

Accordingly, when there is little image motion, it is possible to accurately and efficiently exclude intra-prediction direction candidates with low correlation from known pixel differences.

In the intra prediction device according to the present invention, it is characterized in that the candidate limiting means can adjust the limited number of candidates for the intra prediction direction according to the processing performance of the intra prediction device.

In addition, the present invention can not only be realized as such an internal prediction device, but also can realize the characteristics of this internal prediction device as an internal prediction method in steps, and these steps can also be implemented as a method that can be executed in a computer. program to achieve. Therefore, it goes without saying that such a program can be delivered via a recording medium such as a CD-ROM and a transmission medium such as the Internet. Furthermore, it can also be realized as an image coding device including the above-mentioned intra prediction device.

As can be seen from the above description, according to the intra prediction device related to the present invention, high image quality and high encoding efficiency can be maintained, and calculation processing for intra prediction can be reduced.

That is, although intra prediction is a process of forming a predicted image by copying pixel values from surrounding pixels, in intra prediction, since a plurality of prediction directions are used to select high encoding efficiency, the calculation of the non-selected prediction direction becomes lengthy. Therefore, since it does not contribute to encoding efficiency, it is effective to omit the intra prediction that does not make much difference in encoding efficiency in terms of reducing the amount of encoding calculation. Also, in order to reduce this redundancy in field images of the interlaced scanning method, it is possible to reduce the load on intra prediction by not using a prediction method that does not cause a difference in coding efficiency in advance. In addition, by using the characteristics of the image, the internal prediction processing corresponding to the image can be reduced, and the processing efficiency is higher compared with the existing method.

Therefore, the practical value of the present invention is extremely high in today's widespread use of digital cameras and mobile phones with cameras.

Description of drawings

FIG. 1 is a diagram showing the relationship between a macroblock of 16×16 pixels included in a picture I and each block of 4×4 pixels.

FIG. 2 is a diagram showing directions of 4×4 intra prediction.

FIG. 3 is a diagram showing directions of 16×16 intra prediction.

FIG. 4 is a sequence diagram showing prediction processing performed by a conventional intra prediction device.

FIG. 5 is a functional block diagram showing the overall structure of an image encoding device 1 applied to the intra prediction device according to the embodiment of the present invention.

FIG. 6 is a block diagram showing the functional configuration of the intra prediction device 11 using image feature quantities.

FIG. 7 is a sequence diagram showing detailed processing performed by each part of the intra prediction device 11 .

FIG. 8 is a block diagram showing the functional configuration of the intra prediction device 11 using a picture structure.

FIG. 9 is a block diagram showing the functional structure of the intra prediction means 11 using the aspect ratio.

Fig. 10 is a diagram showing the angular deviation between the original picture and the field for 4x4 intra prediction.

FIG. 11 is a diagram showing the distance relationship between the pixel position of the 16×16 intra prediction field and the pixel position of the original image.

FIG. 12 is a block diagram showing the functional structure of the intra prediction means 11 using motion vectors, prediction modes, and pixel calculation information.

FIG. 13 is a diagram for explaining determination of prediction direction candidates using motion vectors. In particular, Fig. 13(a) is a diagram showing the original images arranged in the coding order, Fig. 13(b) is a diagram showing the coded images arranged in the coding order, and Fig. 13(c) is a diagram showing the coded images in the coding order first A graph of the relationship between images and images of encoded objects.

FIG. 14 is a diagram for explaining determination of prediction direction candidates by prediction mode. In particular, Fig. 14(a) is a diagram showing original images arranged in coding order, Fig. 14(b) is a diagram showing coded images arranged in coding order, and Fig. 14(c) is a diagram showing coded images arranged in coding order first. A diagram of the relationship between an image and an encoding target image.

FIG. 15 is a flowchart showing the operation of the intra-prediction candidate limiting process in which the candidate limiting unit 111 limits intra-prediction candidates using known results, intra-prediction.

Detailed ways

Next, an intra prediction device according to an embodiment of the present invention will be described with reference to the drawings.

FIG. 5 is a functional block diagram showing the overall configuration of an encoding device 1 applied to an intra prediction device according to an embodiment of the present invention.

Encoding device 1 is used to perform internal (intra-picture) predictive coding on picture I, and external (inter-picture) coding on picture P and picture B. As shown in FIG. , a mode switch 13, a transform unit 14, a quantization unit 15, an entropy encoding unit 16, an external prediction unit 17, an encoding control unit 18, and the like. In addition, the external prediction unit 17 can be composed of a mode switch 170, an inverse quantization unit 171, an inverse transformation unit 172, an adder 173, a loop filter 174, a frame memory 175, a dynamic prediction unit 176, and a dynamic supplementary unit 177.

Regarding a picture to be externally coded, the picture to be coded is divided into blocks (macroblocks) of, for example, 16 horizontal by 16 pixels called macroblocks, and the following processing is performed in units of blocks.

A picture to be coded can be converted into a difference image by obtaining the difference between predicted image signals output from motion compensation section 177 in subtracter 12 . The differential image is output to the transformation unit 14 through the mode switching switch 13 .

The difference image is transformed from the time axis to the frequency axis by the transform unit 14, and quantization is further performed by the quantization unit 15, generating a residual signal composed of transform coefficients.

This residual signal is locally decoded by the outer prediction unit 17 . That is, the residual signal is inversely quantized in the inverse quantization unit 171 . Accordingly, inverse transform section 172 generates a residual decoded signal by performing image decoding processing such as inverse frequency transform. The adder 173 generates a reconstructed image signal by adding the residual decoded signal and the predicted image signal. The obtained reconstructed image signal is stored in the frame memory 175 after the block distortion is removed by the loop filter 174 .

On the other hand, the input image signal in units of macroblocks read from frame memory 175 is input to motion prediction section 176 . Here, one or more coded pictures stored in the frame memory 175 are searched for, and the image region closest to the input image signal is detected to determine a motion vector indicating the position and a reference picture index. The index indicates The picture selected at this time is displayed. Detection of motion vectors is performed in units that are further divided into macroblocks. Motion prediction section 176 uses the obtained motion vector and reference picture index to extract the most suitable image area from the coded pictures stored in frame memory 175 to generate a predicted image.

Entropy coding section 16 performs variable-length coding on coded information such as motion vectors, reference picture indexes, and residual coded signals output through the series of processes described above, so that a bit stream with a small amount of data can be output through the coding process.

Although the flow of the above processing is the operation in the case of performing inter-picture predictive coding, it can also be switched to intra-picture predictive coding through the switch 114 and switch 115 .

When performing intra coding, instead of generating a predicted image by motion compensation, the intra prediction device 11 generates a predicted image of the coding target region from coded regions in the same frame, and generates a difference image signal by taking a difference. In addition, this will be described in detail later. As in the case of external encoding, this difference image signal is converted into a residual coded signal in the transform unit 14 and the quantization unit 15, and variable-length coding is performed in the entropy coding unit 16, thereby outputting a bit stream with a small amount of data .

Next, a detailed configuration of the intra prediction device 11 shown in FIG. 5 will be described.

FIG. 6 is a block diagram showing the functional configuration of the intra prediction device 11 using image feature quantities. In addition, the figure also shows an image feature quantity input unit 181 including the encoding control unit 18 .

The image feature input unit 181 of the encoding control unit 18 acquires information added to and from the video stream and information possessed by the image input device as image features, and outputs the acquired image features to the intra prediction device 11 .

The intra prediction means 11 has a candidate defining unit 111 for defining an intra prediction direction candidate applicable to each block of a plurality of pixels constituting an image from among a plurality of features incorporated into the image, and the candidate defining unit 111 defines The intra prediction execution unit 112 performs intra prediction for the intra prediction direction. In addition, the intra-prediction execution unit 112 has a prediction error calculation unit for calculating the difference value between the reference pixel and the coding target pixel for each prediction mode, and calculates the absolute value difference value, a transformation unit with the same function as the transformation unit 14, and a quantization A quantization unit with the same function as unit 15, an inverse quantization unit with the same function as the inverse quantization unit 171, an inverse transformation unit with the same function as the inverse transformation unit 172, and a method for restoring the pixel value of the restored image from the pixel value and difference value of the reference pixel Inverse prediction unit.

Candidate limiting section 111 first limits intra prediction direction candidates based on image feature quantities from image feature quantity input section 181 .

Next, the intra prediction performing unit 112 performs intra prediction on the intra prediction directions defined by the candidate defining unit 111 . That is, the forecast error is calculated and the smallest forecast error is used as the inner forecast.

In intra prediction, there is a correlation between the feature amount of an image and the prediction accuracy of the intra prediction direction.

By using prediction directions with high prediction accuracy derived from image feature classes as prediction direction candidates, it is possible to preferentially omit prediction directions with low prediction accuracy and other prediction directions with no change in prediction accuracy, and reduce the load of intra-prediction processing. The image feature amount mentioned here includes other external factors such as feature amount obtained by calculating images and pixels, and characteristics of image input devices such as imaging devices.

For example, in photography, a photographic lens that rotates the camera horizontally is used. The photographic lens can be detected by a gyro sensor provided in, for example, a camera. In addition, the camera lens can also be detected based on the front and rear motion vectors of the image. Once one frame of the image next to the photographing lens is taken out, a frame with strong inter-pixel correlation is formed in the lateral direction. Based on the correlation between pixels, the prediction accuracy of the intra prediction direction can be estimated, and intra prediction direction candidates with low prediction accuracy can be subtracted in advance.

In addition, when the machine is set to encode at a low bit rate, in the case of encoding at a low resolution, and when using a similar prediction direction such as prediction mode 0 (horizontal) and prediction mode 6 (horizontal-down) case, there is no difference in prediction error between the two modes, and no difference in coding performance. Here, the prediction mode 0 (horizontal) and the like in which the directions of the candidate prediction directions are greatly different are preferentially predicted. That is, prediction mode 6 is preferentially deleted.

In addition, the prediction direction can be determined according to the motion vector. That is, when intra prediction is performed, motion vectors of past or future frames are used as image feature quantities. There is a high probability of highly correlated pixel values in the direction of this motion vector. Thus, intra-prediction directions that greatly deviate from the direction of the motion vectors can be omitted. Here, the past or future frame motion vector may be a motion vector of the entire screen, or may be a motion vector of a specific area within the screen.

In addition, it is possible to omit an intra prediction direction that greatly deviates from the direction of the motion vector near the adjacent MB of the coded MB from the intra prediction direction candidates of the coded MB.

Furthermore, in coding MB, when a motion vector has already been obtained, if the motion vector is used as a feature quantity, it is possible to further select high-precision intra prediction direction candidates.

In H.264, only intra prediction is performed on a picture I, and both intra prediction and extrinsic prediction are performed on a P or B picture, thereby performing encoding in a coding mode with few residual signals. Therefore, intra prediction can be performed on all pictures.

By using these, the candidates for the prediction direction of the coding target picture are limited. That is, candidate limiting section 111 can limit candidates for intra prediction directions based on results obtained from original images or encoded images that are encoded earlier in order. Accordingly, it is also possible to easily limit candidates for intra-prediction directions.

Specifically, first, the prediction direction of intra-picture prediction and the MB position in the picture encoded first are recorded in memory in advance. Next, intra prediction of the same MB position as that of the encoding-processed picture is performed using the intra-picture MB position and prediction direction recorded in the memory. For example, when the entire screen does not move, there is a high possibility that consecutive pictures in display order are in the same direction with respect to the prediction direction of the intra prediction of the spatial position MB. For example, in the case of encoding in the display order I, B1, B2, P... structure, the prediction direction of a certain MB in the B1 picture is the same as the encoding result of the MB at the same position as the picture I or picture P. The odds are high. Therefore, the prediction directions that are the coding results of the picture I and the coding results of the picture P are taken as candidates for the intra prediction direction of the picture B1. Similarly, picture B2 has the intra prediction directions of picture B1 and picture P as candidates. In the case of selecting the common extrinsic prediction of the picture B1 and the picture P, the prediction candidate of the picture B2 is set to 0, and only the extrinsic prediction may be performed.

Here, although MBs at the same position in a picture are mentioned, the prediction directions of adjacent MBs may be included in the candidates. That is, in the above example, by adding MBs at the same position as the picture B1 or the picture P as the prediction candidates of the picture B2, the MB prediction directions of up, down, left, and right can be used as candidates.

FIG. 7 is a sequence diagram showing detailed processing performed by each unit of the intra prediction device 11 .

Once the picture I is input, the candidate limiting unit 111 first limits the candidates for the intra prediction direction based on the image feature quantity from the image feature quantity input unit 181 . This definition can apply to either or both blocks and macroblocks.

Next, the intra prediction performing unit 112 performs intra prediction for the intra prediction directions defined by the candidate defining unit 111 . More specifically, the prediction error (sum of absolute value differences) is calculated for the prediction mode defined for block 0 . That is, only the difference value and the absolute value difference sum between the reference pixel and the coding target pixel are calculated for the limited prediction mode. Thus, the intra prediction performing unit 112 uses the one with the smallest prediction error as the intra prediction. That is, in the case of the absolute value difference and the minimum prediction mode 0, by transforming, quantizing, inverse quantizing, inverse transforming, and inversely predicting the difference value between the reference pixel of the block 0 related to the prediction mode 1 and the pixel to be encoded, The pixel value of the restored image of block 0 is restored from the pixel value of the reference pixel and the difference value.

Once the pixel values of the restored image of block 0 are restored, the intra prediction execution unit 112 repeats the same processing as above, thereby restoring the pixel values of the restored images of blocks 1 to 15 .

Once the restored image of the block 15 is restored, only the difference value and the absolute value difference sum of the reference pixel and the pixel to be encoded are calculated in the imported prediction mode for the macroblock. Thus, the intra prediction performing unit 112 uses the one with the smallest prediction error, that is, for example, the total number of absolute value difference sums of the respective blocks and the smaller absolute value difference sum of the macroblocks as the intra prediction. That is, in the case where the total number of absolute value difference sums for each block is small, the intra prediction performing unit 112 outputs the difference value for each block. On the contrary, in the case where the absolute value difference sum of the macroblock is small, the intra prediction performing unit 112 outputs the difference value for the macroblock.

Therefore, when the prediction mode is limited, the load on the intra-prediction execution unit 112 can be greatly reduced, and the calculation performance of the intra-prediction can be improved.

In addition, such an internal prediction device 11 is not only a hardware configuration, but can also be realized by means such as a CPU, a memory, and a program.

Next, the case where candidates are limited by combining image feature values will be described in order.

First, a case where the image feature quantity is the pixel aspect ratio will be described.

FIG. 8 is a block diagram showing the functional configuration of the intra prediction device 11 using the pixel aspect ratio as an image feature quantity. In addition, in this figure, the pixel aspect ratio detection section 183 included in the image feature quantity input section 181 incorporating the encoding control section 18 is also shown. In addition, the same reference numerals are given to the structural parts corresponding to the case of FIG. 6, and their detailed descriptions are omitted.

In the field structure, it is possible to record images of horizontal pixels using every other pixel. Thereby, a change occurs in the direction on the field and the direction of the original image. Likewise, when the pixel aspect ratio representing the horizontal and vertical distance ratio between adjacent pixels is changed, the original image direction and the image direction of the pixel aspect ratio are changed. The pixel aspect ratio detection unit 183 detects this change. Corresponding to the change of the pixel aspect ratio in the direction of the original image, the candidate limiting unit 111 can define appropriate prediction direction candidates, thereby improving the calculation performance of the intra prediction.

A specific example in the case of changing the pixel aspect ratio is described below. Since the image of half D1 maintains an interval between pixels in the horizontal direction, the direction on half D1 and the direction on the original image change. Therefore, there are prediction directions that do not contribute to coding efficiency as in the case of the field structure. Specifically, they are prediction mode 6 and prediction mode 8 shown in FIG. 2 . These, once restored to the orientation of the original image, are close to the horizontal orientation of prediction mode 0, and make no difference in coding efficiency. Therefore, by omitting the prediction direction, it is possible to reduce calculation processing. In addition, the prediction direction of the intra prediction can be corrected in the prediction direction corresponding to the pixel aspect ratio, so that the performance in all prediction directions can also be improved.

In addition, such an internal prediction device 11 is not only constituted by hardware, but may be realized by means such as a CPU, a memory, and a program.

In addition, in the intra prediction device 11, horizontal and vertical distances on the original image can be used as a method of defining appropriate prediction direction candidates according to the pixel aspect ratio.

Take the reciprocal of each distance as the confidence of the prediction. For example, in the field structure, the distance on the original image sets the horizontal distance to "1", and sets the vertical distance to "2", which is twice the horizontal distance. Thus, the reliability levels are "1" and "1/2".

Since this is the case, the determination of the vertical direction is greater than that of the horizontal direction, so the prediction results are along this tendency. Therefore, as candidates for the predicted direction, candidates close to the horizontal direction are preferentially adopted. That is, the load can be reduced by preferentially deleting the vertical direction and directions close to this direction.

In addition, what is described here is not limited to a device, but may be realized by means such as a program.

In addition, pixel aspect ratio information can be acquired from a camera or video recorder, and intra prediction can be performed using the information.

For example, in the case where the imaging element of the camera is not square, there will be a difference in the vertical distance and the horizontal distance on the original image. In addition, by changing the screen aspect ratio in a video recorder or the like, a difference is also generated in the vertical distance and the horizontal distance of the original image in the case of recording.

In addition, some encoding methods perform internal prediction by adding information inside and outside the video stream, such as obtaining information from the VUI parameter aspect ratio idc of H.264.

In addition, what is described here is not limited to a device, but may be realized by means such as a program.

Next, a case where the image feature quantity is a picture structure will be described.

FIG. 9 is a block diagram showing the functional configuration of the intra prediction device 11 using a picture structure as an image feature value. However, this figure also shows a picture structure determination unit 182 included in the image feature value input unit 181 of the encoding control unit 18 . In addition, the same reference numerals are assigned to the components corresponding to those in the case of FIG. 6 , and their detailed descriptions are omitted.

The field image of the interlaced scanning method is formed by maintaining a pixel interval of one original image in the vertical direction. Therefore, although the continuity of pixels in the horizontal direction does not change compared with the frame image of the progressive scan method, the continuity of pixels in the vertical direction is lower than that of the frame image. The level of continuity is determined by the picture structure determination unit 182 .

By using this characteristic, it is easier for candidate limiting section 111 to select prediction in the horizontal direction than prediction in the vertical direction. That is, prediction in the vertical direction (4×4 prediction mode 0 and 16×16 prediction mode 0) is preferentially excluded from candidates.

However, what is described here is not limited to a device, but may be realized by means such as a program.

In addition, in the interlaced scanning method, once the 4×4 intra prediction direction is converted into the direction on the original image, as shown in FIG. 10 , no difference occurs in the prediction method.

In a certain field image in the interlaced scanning mode, straight lines in 8 directions as shown in FIG. 10 are introduced.

As shown in FIG. 10( a ), an interlaced field image has a structure in which pixel intervals are maintained for each line, and the distance is compressed to 1/2 in the vertical direction. On the contrary, once it is repositioned at the pixel position on the original image as shown in Fig. 10(b), the vertical distance becomes twice the horizontal distance without stretching the spacing. Therefore, on the original image, the angles of the straight lines in the 8 directions vary respectively.

Particular attention should be paid to the direction of prediction mode 5 (vertical-right), prediction mode 7 (vertical-left). These directions are very close to those of prediction mode 0 (vertical).

Therefore, in 4×4 intra prediction, the pixel values predicted in the direction of prediction mode 5 (vertical-right), prediction mode 7 (vertical-left), prediction mode 0 (vertical) are very similar, so as to be compared with the original Differences are less likely to occur in poor prediction errors of images. Similarly, the same tendency appears in the directions of prediction mode 3 (diagonal-bottom-left) and prediction mode 4 (diagonal-bottom-right).

Using this feature, if the prediction candidates are limited in advance, the internal prediction process can be simplified without degrading the image quality.

In this case, the picture structure judging section 182 judges whether the picture structure is an interlaced field image or a progressive frame image, and outputs the judgment result to the intra prediction device 11 .

If it is a field image, the candidate limiting unit 111 of the intra prediction device 11 sends the intra prediction candidates for the field to the intra prediction execution unit 112, and the intra prediction candidates for the field omit several prediction candidates as shown in FIG. 9 Candidates for intra-field prediction. The intra prediction execution unit 112 determines the prediction direction by performing prediction calculation processing on the limited intra prediction direction candidates.

In addition, what is described here is not limited to a device, and may be realized by means such as a program.

Furthermore, generally, when prediction is performed using pixels of adjacent MBs, the correlation between pixels becomes lower than the distance from the adjacent MBs.

Therefore, as the distance between pixels on the original image becomes longer, there is a greater possibility that the prediction error will become larger.

Where there is a difference between the horizontal distance and the vertical distance, there is also a difference between the prediction error for prediction mode 0 (vertical) and the prediction direction through prediction mode 1 (horizontal) for the intra 16x16 prediction.

When the picture structure is interlaced, as shown in FIG. 11 , the vertical distance between pixels on the original image is twice the horizontal distance, the pixel correlation in the vertical direction becomes low, and the prediction error increases.

In interlaced field images, prediction mode 0 (vertical) prediction related to 16×16 intra prediction can be preferentially omitted.

Likewise, in the prediction of prediction mode 2 (average) based on the average prediction of the right end pixel value of the adjacent right MB and the lower end pixel value of the adjacent upper MB, pixels with low vertical correlation are used for prediction pixel calculation so that The prediction error becomes large, and the prediction mode 2 (average) related to 16×16 intra prediction can be preferentially omitted.

In addition, what is described here is not limited to a device, and may be realized by means such as a program.

In addition, the determination of whether the image recording method is the interlaced scanning method or the progressive scanning method may be as follows. Hereinafter, when recording is performed by designating an interlaced scanning method or a progressive scanning method, the recording method and reproduction time are referred to as a reproduction method.

There are cases selected by the user and cases determined by specification parameters of the camera, video recorder, display device, and the like.

For example, considering the selection by the user, if the camera supports both interlaced scanning and progressive scanning, the user selects the recording method by combining the playback environment such as a display device and a playback device configured by the user. If the television as the display device is an interlaced scanning method, by selecting the interlaced scanning method at the time of encoding in a video recorder as a camera device, the intra prediction direction which does not contribute to the encoding efficiency can be omitted.

For example, considering the case of determining by the specification parameters of the camera, if the camera supports only the interlaced scanning method, the encoding efficiency can be improved by omitting the intra prediction direction when shooting in the interlaced scanning method.

In addition, considering the situation determined by the specification parameters of the video recorder, when encoding from another image input device in a video recorder, etc., if it is judged that the input of the image device is an interlaced scanning method, then by omitting the intra prediction direction, it is possible Improve coding efficiency.

The inputs from other image input machines described here are supplemented below.

In the case of encoding using another encoding method, it can be determined whether it is an interlaced scanning method based on the rules of the encoding method. That is, the determination may be made based on header information or the like.

Furthermore, considering the fact that it is determined by the specification parameters of the display device, the encoding efficiency can be improved by combining the playback method of the display device connected to the video recording device and omitting intra prediction when recording in the interlaced scanning method.

In addition, what is described here is not limited to a device, and may be realized by means such as a program.

In addition, the so-called H.264 intra-prediction picture-level adaptive means that for each picture, by comparing the intra-prediction results of the relevant frame structure and the intra-prediction results of the relevant field structure, the one with the best coding performance is used as the prediction result. Way. When performing prediction as the field structure, the candidate limiting section 111 that limits candidates for the prediction direction in the picture structure is used.

Since it is already known that the picture structure judging unit 182 uses picture-level adaptation in encoding, in the case of using this method, the candidate limiting unit 111 can reduce the number of predictors when encoding with a field structure, that is, encoding with an interlaced scanning method. alternate direction.

In addition, the so-called H.264 intra-prediction macroblock-level adaptation refers to the method of comparing the intra-prediction results related to the frame structure and the intra-prediction results related to the field structure in units of MB, and taking the one with the best coding performance as the prediction result. . When the field structure is predicted, the candidate limiting section 111 for limiting candidates of the prediction direction in the picture structure is used. Since it is already known that the picture structure judging unit 182 uses macroblock-level adaptation in encoding, when this method is used, when encoding is performed with a field structure, that is, when encoding is performed in an interlaced scanning manner, the candidate limiting unit 111 can reduce the number of predictions. alternate direction.

In addition, what is described here is not limited to a device, and may be realized by means such as a program.

Next, a case will be described where the image feature value is obtained from an original image or an encoded image earlier in the encoding order.

FIG. 12 is a block diagram showing the functional structure of the intra prediction means 11 using motion vectors, prediction modes, and pixel calculation information as the above results. In addition, in this figure, motion vector input section 184 , prediction mode input section 185 , and pixel calculation information input section 186 included in image feature quantity input section 181 of encoding control section 18 are also shown. In addition, the same reference numerals are assigned to components corresponding to those in the case of FIG. 6 , and detailed descriptions thereof are omitted.

As mentioned above, H.264 only performs intra-prediction on picture I, and implements intra-prediction/external prediction on picture P or B, and encodes in a coding mode with less residual signal.

Therefore, for an original image or an encoded image whose encoding order is earlier, results such as adjacent pixel difference and intra prediction direction are known. In addition, when the encoded pictures preceding in the encoding order are the picture P and the picture B, the motion vectors of the respective macroblocks are known. Furthermore, depending on the characteristics of images, there are many cases in which the correlation between an original image or an encoded image and an encoding target image is extremely high.

Thus, even by using the result obtained for the original picture or the coded picture that is earlier in the coding order, that is, even if the candidates for the intra prediction direction are defined using known motion vectors, intra prediction sums, and adjacent pixel differences, it is possible to advance Processing with poor coding efficiency and processing with no difference in coding efficiency is omitted.

Therefore, in this embodiment, as shown in FIG. 12 , in the image feature quantity input unit 181, a motion vector, a prediction mode, and a prediction mode respectively used to input a result obtained for an original image or an encoded image earlier in the encoding order can be set. A motion vector input unit 184 , a prediction mode input unit 185 , and a pixel calculation information input unit 186 for the direction of and adjacent pixel differences (also referred to as pixel calculation information). Therefore, image feature quantity input section 181 can selectively output the motion vector, prediction mode direction, and adjacent pixel difference from motion vector input section 184, prediction mode input section 185, and pixel calculation information input section 186 to candidate definition section 111. to form.

Next, when a motion vector is output from the motion vector input section 184, determination of prediction direction candidates based on the motion vector will be described.

Fig. 13 is a diagram for describing the determination of prediction direction candidates by motion vectors. In particular, Fig. 13(a) is a diagram showing the original images arranged in the coding order, Fig. 13(b) is a diagram showing the coded images arranged in the coding order, and Fig. 13(c) is a diagram showing the coded images arranged in the coding order first. A diagram of the relationship between an image and an encoding target image.

Here, it is shown that a coded picture immediately preceding a coding target picture is used as a coded picture having an earlier coding order. This is because the correlation between the previous original picture or coded picture and the picture to be coded becomes high, and it is possible to precisely define intra prediction direction candidates.

Therefore, this coded image is composed of macroblocks MB1 to MB9 as shown in the figure, and the motion vectors of the respective macroblocks MB1 to MB9 are described as MV1 to MV9. In addition, the description will be made by taking the encoding target macroblock as α shown in the figure as an example.

In the case of performing intra-prediction on macroblock α, the candidate limiting unit 111 uses the macroblock MB9 having the same spatial position in the previous coded picture as a unit area, refers to the motion vector MV9 of the unit area, and based on the direction of the motion vector MV9, Candidates for the internal prediction direction are limited.

Specifically, the motion vector MV9 has many components in the horizontal direction and very few components in the vertical direction. That is, the horizontal component in the direction of the motion vector MV9 is dominant, and has a high correlation with the horizontal direction of the current macroblock α. Therefore, the prediction direction in the vertical direction by candidate limiting section 111 is excluded from the candidates. That is, candidate limiting section 111 omits patterns 0, 5, and 7 from motion vector MV9.

Therefore, in the case of moving images, the accuracy of the motion vector of the macroblock to be coded can be improved, and intra prediction direction candidates with low correlation can be accurately and effectively excluded from known motion vectors.

In addition, by using the area around the macroblock MB9, that is, the macroblocks MB5, MB6, and MB8 as a unit area, candidates for the intra prediction direction can be specified based on the motion vector directions of the unit area. That is, candidates for the direction of intra prediction may be limited based on the directions of the motion vectors of the synthesized motion vectors MV5 , MV6 , MV8 , and MV9 . Even in this case, since the directions of the composite motion vectors are obtained by decomposing and adding the motion vectors MV5, MV6, MV8, and MV9 into horizontal and vertical directions, respectively, the horizontal direction component is dominant. Therefore, candidate limiting section 111 can exclude the prediction direction in the vertical direction from the candidates. That is, candidate limiting section 111 omits patterns 0, 5, and 7 from motion vectors MV5, MV6, MV8, and MV9.

In addition, it is also possible to specify intra-prediction candidates by referring to which direction the motion vector of the entire preceding coded picture is directed. That is, the entire macroblocks MB1 to MB9 may be used as a unit area, and the intra prediction direction may be determined based on the motion vector direction of the unit area.

Next, a description will be given regarding the case where the prediction mode is output from the prediction mode input unit 185 and the prediction direction candidates are determined by the prediction mode.

FIG. 14 is a diagram for describing determination of prediction direction candidates by prediction mode. In particular, Fig. 14(a) shows the original images arranged according to the coding order, Fig. 14(b) shows the coded pictures arranged according to the coding order, and Fig. 14(c) shows the coded pictures arranged in the coding order first and the coding target picture. diagram of the relationship between. Furthermore, as shown in the figure, this coded image is described as being composed of macroblocks MB1 to MB9. In addition, the coding target macroblock is described as α shown in the figure.

Here, in the case where the encoded image is the picture I, for the macroblocks employed in the intra prediction, the results of the defined prediction modes are preserved. Thus, in this case the direction of the defined prediction mode can be used. On the other hand, when the coded image is the picture P or B, the result of intra prediction and the prediction mode are usually discarded. However, in this embodiment, the results and prediction modes at the time of intra prediction are described as the results configured to be temporarily stored in the memory.

FIG. 15 is a flowchart showing the operation of the intra-prediction candidate limiting process in which the candidate limiting unit 111 limits intra-prediction candidates using known results, intra-prediction.

When intra prediction is performed on macroblock α, candidate limiting section 111 first judges whether or not a reference coded picture preceding in coding order is picture I (S11). In addition, a coded picture preceding the coding target picture may be used as a coded picture having a preceding coded picture. This is because the correlation between the previous original picture or coded picture and the picture to be coded becomes high, and the intra prediction direction candidates can be precisely limited.

When the determination result is picture I (Yes in S11), candidate limiting section 111 limits intra prediction candidates using the prediction result of picture I (S12), and ends the intra prediction candidate limiting process. That is, the candidate limiting section 111 takes the macroblock MB9 having the same spatial position in the previous coded image as a unit area, refers to the direction of the intra prediction mode of the unit area, and predicts the macroblock used for the intra prediction based on the limited prediction In terms of the mode direction, candidates for the intra-prediction direction are limited. Specifically, if the direction of the limited prediction mode is, for example, the horizontal direction, the probability that the intra prediction direction of the macroblock α to be coded is the horizontal direction is high. Therefore, candidate limiting section 111 excludes the prediction direction in the vertical direction from the candidates.

Therefore, even if there is motion in the image, in the case of less motion in the image, the accuracy of the prediction direction of the macroblock to be coded can be improved, and the direction of the known prediction mode can be accurately and efficiently Candidates for internal prediction directions with low correlation are excluded.

Also, by using the area around the macroblock MB9 corresponding to the current macroblock α, that is, the macroblocks MB5, MB6, and MB8 as the unit area, the intra prediction direction can also be determined based on the prediction mode direction of the unit area. alternate.

On the other hand, when it is not the picture I (No in S11), the candidate limiting section 111 limits the intra prediction candidates using the prediction result in the case of intra encoding (S13), and ends the intra prediction candidate introduction process.

Therefore, even when the picture to be coded earlier in the coding order is not the picture I, no matter how the picture moves, it is possible to accurately, easily and efficiently exclude candidates for intra prediction directions with low correlation from known intra prediction directions.

Next, regarding the case where the difference between adjacent pixels, that is, the pixel calculation information is output from the pixel calculation information input section 186, the determination of the prediction direction candidates by the pixel calculation information will be described.

Here, the so-called pixel calculation information means any one of the adjacent pixel difference of the original image and the adjacent pixel difference of the coded image included in the image feature, which is encoded first. Such pixel calculation information is calculated and known in advance. Therefore, by temporarily storing the result in the memory so that it can be reused or recalculated, the load on the candidate limiting unit 111 can be reduced. In addition, here, the description will be made with a structure in which the result of the intra prediction and the pixel calculation information are temporarily held in the memory.

Candidate limiting section 111 limits intra-prediction candidates based on adjacent pixel differences of an original image or encoded image that is included in the image feature and is encoded earlier. In addition, an original image or a coded image preceding the encoding target image may be used as an original image or an encoded image preceding the encoding order. This is because the correlation between the previous original picture or coded picture and the picture to be coded increases, so that candidates for the intra prediction direction can be precisely defined.

Furthermore, the candidate limiting unit 111 uses a macroblock with the same spatial position in the previous original image or coded image as a unit area, refers to the pixel calculation information of the unit area, and based on the pixel calculation information of the macroblock used in the intra prediction, The macroblock α of the encoding target image defines candidates for intra prediction directions. Specifically, intra-prediction candidates are limited based on the ratio of the difference between adjacent pixels in the horizontal direction of the unit area and the difference between adjacent pixels in the vertical component. Thereby, it can be known which of the horizontal direction and the vertical direction has the highest correlation, that is, the direction with the highest correlation. Therefore, when the amount of motion of the image is small, it is possible to accurately and efficiently exclude intra-prediction direction candidates with low correlation from the known pixel calculation information.

In the above-described embodiment, the candidate limiting unit may adjust the limited number of intra prediction direction candidates according to factors different from the characteristics of the image data. For example, the limited number can be adjusted from intra prediction direction candidates according to the processing performance of the intra prediction device.

In addition, what is described here is not limited to a device, and may be realized by means such as a program.

In addition, although the above-mentioned embodiment has described the case where the image is a moving image, the object of the intra prediction can also be the picture I, and the intra prediction according to the present invention can also be applied to still images such as JPEG, and further compression coding can be performed. data.

If the intra-prediction device according to the present invention is used, the image encoding of H.264 can be realized with less calculation load, and it can be applied to portable telephones equipped with video cameras and cameras in addition to personal computers, HDD recorders, DVD recorders, etc. . In addition, it is also applicable to an encoding device including the intra prediction device.

Claims (24)

1. An internal forecasting device, characterized in that it comprises: candidate limiting means for defining candidates of intra-prediction directions suitable for a plurality of pixel blocks constituting the image data from among a plurality of intra-prediction directions by adapting to a characteristic of the image data, and intra-prediction performing means for performing intra-prediction using the intra-prediction direction defined by the candidate defining means. 2. The intra-prediction device according to claim 1, wherein said candidate limiting means limits candidates for an intra-prediction direction based on a pixel aspect ratio characteristic of said image data. 3. The intra prediction device according to claim 2, wherein the aspect ratio is obtained from any one of an image signal input device and information obtained from inside or outside the video stream. 4. The intra prediction device according to claim 2, wherein the candidate limiting means uses the horizontal and vertical distances between adjacent pixels in the original image according to the pixel aspect ratio as the horizontal and vertical distances of the intra prediction. The reliability of the prediction in the vertical direction is used to limit the candidates for the internal prediction direction. 5. The intra-prediction device according to claim 1, wherein said candidate limiting means limits intra-prediction based on picture structure information that is characteristic of said image by indicating which one is interlaced scanning and progressive scanning. direction alternate. 6. The intra-prediction device according to claim 5, wherein the picture structure information is obtained from any one of image signal input equipment and information obtained from inside or outside the video stream. 7. The intra prediction device according to claim 5, wherein when the picture structure information indicates interlaced scanning, the candidate limiting means preferentially excludes the vertical direction included in the intra prediction direction from the candidates. -right direction and vertically-left direction. 8. The intra prediction device according to claim 5, wherein when said picture structure information indicates interlaced scanning, said candidate limiting means preferentially excludes vertical- Right Direction, Vertical-Left Direction, Diagonal-Down-Right Direction, and Diagonal-Down-Left Direction. 9. The intra prediction device according to claim 5, wherein when the picture structure information indicates interlaced scanning, the candidate limiting means preferentially excludes the average value included in the intra prediction direction from the candidates. direction. 10. The intra prediction device according to claim 5, wherein when the picture structure information indicates interlaced scanning, the candidate limiting means preferentially excludes the average value included in the intra prediction direction from the candidates. direction and vertical direction. 11. The intra-prediction device according to claim 5, wherein said candidate limiting means limits candidates for intra-prediction directions based on picture structure information when encoding is performed by picture-level adaptation related to a video compression method. 12. The intra-prediction device according to claim 5, wherein the picture structure information is any one of field structure and frame structure, and when encoding is performed by macroblock-level adaptive coding related to a moving picture compression method, in When performing prediction by switching between a pair of macroblocks in a field structure and a pair of macroblocks in a frame structure, the candidate limiting means limits candidates for intra prediction based on information on whether the macroblock pair has a field structure or a frame structure. 13. The intra prediction device according to claim 1, wherein said candidate limiting means limits candidates for an intra prediction direction based on a result obtained from an original picture or a coded picture having a preceding encoding order. 14. The intra prediction device according to claim 13, wherein an original picture or a coded picture preceding the picture to be coded can be used as the original picture or coded picture preceding the coding order. 15. The intra-prediction device according to claim 13, wherein the result is a motion vector obtained from a coded picture earlier in coding order, and the candidate limiting means limits the intra-prediction based on the direction of the motion vector direction alternate. 16. The intra prediction device according to claim 15, wherein, as the motion vector, a position indicating the same space as the macroblock to be coded, a position including the surrounding of the same space position, and an entire picture can be used. Any one of the dynamic vectors of the unit field. 17. The intra prediction device according to claim 13, wherein said result is an intra prediction direction obtained from a coded picture earlier in coding order, and said candidate limiting means limits intra prediction based on said intra prediction direction direction alternate. 18. The intra prediction device according to claim 17, wherein as the intra prediction direction, a macroblock at the same spatial position as the macroblock to be coded and a macroblock including a position around the same spatial position can be used. Internal forecast direction. 19. The intra-prediction device according to claim 13, wherein said result is an adjacent pixel difference of said original image or encoded image earlier in encoding order, and said candidate limiting means defines Internal prediction direction candidate. 20. The intra prediction device according to claim 1, wherein said candidate limiting means adjusts the limited number of candidates for said intra prediction direction according to the processing performance of said intra prediction device. 21. An internal forecasting method, characterized by comprising: a candidate defining step of defining candidates of intra prediction directions suitable for a plurality of pixel blocks constituting the image data from a plurality of intra prediction directions by adapting to characteristics of the image data, and An intra prediction executing step of executing intra prediction using the intra prediction direction candidates defined in the candidate defining step. 22. The intra prediction device according to claim 21, wherein the limited number is adjusted from the intra prediction direction candidates according to the processing performance of the intra prediction device executing the intra prediction method. 23. A computer program, characterized in that it is a computer program recorded on a computer-readable medium for performing internal prediction in a computer, comprising: A computer-executable program code for executing, in a computer, a candidate defining step of defining candidates of intra prediction directions applicable to a plurality of pixel blocks constituting the image data from among a plurality of intra prediction directions by adapting to characteristics of the image data ,and A computer-executable program code for executing in a computer an intra prediction performing step of performing intra prediction by using the intra prediction direction candidates defined by the candidate defining step. 24. The computer program according to claim 23, wherein the limited number is adjusted from the candidates of the intra prediction direction according to the processing performance of the computer.

Images