CN101218829A - Dynamic image encoding device, dynamic image encoding method, dynamic image encoding program, dynamic image decoding device, dynamic image decoding method, and dynamic image decoding program - Google Patents
Dynamic image encoding device, dynamic image encoding method, dynamic image encoding program, dynamic image decoding device, dynamic image decoding method, and dynamic image decoding program Download PDFInfo
- Publication number
- CN101218829A CN101218829A CNA2006800246255A CN200680024625A CN101218829A CN 101218829 A CN101218829 A CN 101218829A CN A2006800246255 A CNA2006800246255 A CN A2006800246255A CN 200680024625 A CN200680024625 A CN 200680024625A CN 101218829 A CN101218829 A CN 101218829A
- Authority
- CN
- China
- Prior art keywords
- mentioned
- piece
- unit
- dynamic image
- decoding
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
It is possible to effectively encode dynamic image data by deciding a prediction signal without using a motion vector. A dynamic image encoding device (100) includes: a region division unit (101) for dividing a frame image constituting dynamic image data as an object to be encoded into a plurality of regions; an encoding unit (104) for encoding images of the respective regions; an inverse conversion unit (105) and an addition unit (106) for generating a reproduction image of the encoded image; a storage unit (107) for storing reproduced image; a prediction generation unit (108) for searching for a region of an image adjacent to the region of the image to be encoded in a predetermined positional relationship and having a high correlation with the reproduced image of the template region as a part of the reproduced image, from the reproduced image, and deciding a prediction signal according to the searched region and the positional relationship; and a subtraction unit (102) for generating a differential signal between the prediction signal and the encoding object image as a signal for encoding.
Description
Technical field
The present invention relates to moving picture encoding device, motion image encoding method, dynamic image encoding, dynamic image decoding device, dynamic image decoding method and moving image decoding program.
Background technology
The international standard standard that the coding of dynamic image data uses the regulation of ITU-T (standardization department of international telecommunication union telecommunication) for example i.e. moving picture encoding mode H.264.The technology relevant with moving picture encoding mode H.264 for example done open in following non-patent literature 1 grade.
In above-mentioned H.264 moving picture encoding mode, by carrying out motion compensation and motion vector being encoded, cut down the redundancy of dynamic image interframe, cut down amount of information.In the method, can use a motion vector to carry out the zone of motion compensation, also can tackle trickle motion, improve the precision of prediction of inter prediction by dwindling.On the other hand, because the quantity of motion vector increases, it is huge that amount of information becomes, thereby be necessary this is suppressed.In the past, as the method that in the quantity that reduces motion vector, improves precision of prediction, the method for following patent documentation 1 record had been proposed.
Patent documentation 1: Japanese kokai publication hei 2-62180 communique
Non-patent literature 1: wild he, " the H.264/AVC Books イ of religion section Application プ レ ス Standard Quasi religion Books シ リ one ズ of section ", イ Application プ レ ス ネ Star ト PVC ジ ネ ス カ Application パ ニ 1,2004 year
Yet, in the method for above-mentioned patent documentation 1 record, also need motion vector not change, have to motion vector is encoded.According to this situation, need further to improve the dynamic image data coding method of compression efficiency.
Summary of the invention
The present invention In view of the foregoing finishes just, and the present invention can provide a kind of and carry out situation that inter prediction encodes specific energy moving picture encoding device, motion image encoding method, dynamic image encoding, dynamic image decoding device, dynamic image decoding method and the moving image decoding program of more effectively encoding mutually with using motion vector.
Moving picture encoding device of the present invention, it is that unit encodes to dynamic image data with the piece, it is characterized in that this moving picture encoding device has: cutting unit, its two field picture that will constitute dynamic image data is divided into a plurality of coded object pieces; Coding unit, it is encoded to the coded object piece; The reproduced image generation unit, its reproducing signal that generates the coded object piece is a decoding block; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generation unit, its use according to the position relation of regulation and coded object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data by cell stores generates, the prediction signal that generates the coded object piece is promptly predicted piece; To be unit with the pixel deduct the difference signal that the prediction piece generates this coded object piece from the coded object piece to coding unit is difference block, and this difference block is encoded; The reproducing signal that the reproduced image generation unit generates by the difference block behind the coding unit coding is the decoded difference value piece, and is that unit is with this decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.
In moving picture encoding device of the present invention, use according to the position relation of regulation and coded object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data generates, the prediction signal that generates the coded object piece is promptly predicted piece.Use this prediction piece to encode.That is,, do not use motion vector promptly to predict piece by the generation forecast signal, can carry out efficient coding according to moving picture encoding device of the present invention.
Moving picture encoding device of the present invention, it is that unit encodes to dynamic image data with the piece, it is characterized in that this moving picture encoding device has: cutting unit, its two field picture that will constitute dynamic image data is divided into a plurality of coded object pieces; Coding unit, it is encoded to the coded object piece; The reproduced image generation unit, its reproducing signal that generates the coded object piece is a decoding block; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; Retrieval unit, it is from the dynamic image data of reproduction by cell stores, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal with the position relation of regulation and coded object piece in abutting connection with and belong to the dynamic image data of reproduction by cell stores; And prediction signal decision unit, it is based on the position relation of pixel groups that is retrieved by retrieval unit and regulation, according to promptly predicting piece by the prediction signal of reproducing dynamic image data decision coded object piece of cell stores; To be unit with the pixel deduct the difference signal that the prediction piece generates this coded object piece from the coded object piece to coding unit is difference block, and this difference block is encoded; The reproducing signal that the reproduced image generation unit generates by the difference block behind the coding unit coding is the decoded difference value piece, and is that unit is with this decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.According to this structure, retrieve and the high pixel groups of template correlation from reproducing the dynamic image data, and decide the prediction piece according to the position relation of pixel groups that retrieves and regulation.Therefore, piece can be determined to predict reliably, the present invention can be implemented reliably.
Preferably, moving picture encoding device also has: estimation unit, and it compares with template with by the dynamic image data of reproduction of cell stores, according to the spatial continuity of the image of this comparative result estimated coding object piece; And setup unit, it is according to the spatial continuity of the image that is estimated by estimation unit, further partition encoding object piece, and the coded object piece after this is cut apart is set at new coded object piece, and sets the template at this new coded object piece.According to this structure owing to can suitably select to predict the size of piece according to the spatial continuity of reproduced image data, even thereby change in the coding of violent dynamic image data at amount of exercise, also can improve code efficiency.And, owing to change the shape and size of template zone and estimation range according to signal characteristic, thereby can improve the estimated performance of prediction signal.
Dynamic image decoding device of the present invention, to be unit with the piece with the coded data of dynamic image data be reproduced as reproduces dynamic image data for it, it is characterized in that, this dynamic image decoding device has: decoding unit, and its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded; The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated the decoder object piece by the coded data after the decoding unit decodes; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generation unit, its use according to the position relation of regulation and decoder object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data by cell stores generates, the prediction signal that generates this decoder object piece is promptly predicted piece; The difference signal that decoding unit generates the decoder object piece is the decoded difference value piece; The reproduced image generation unit is that unit is with decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.
In dynamic image decoding device of the present invention, can with above-mentioned moving picture encoding device generation forecast piece in the same manner, and dynamic image decoded.That is,, can be correctly decoded carried out the dynamic image data after the efficient coding by above-mentioned moving picture encoding device according to dynamic image decoding device of the present invention.
Dynamic image decoding device of the present invention, to be unit with the piece with the coded data of dynamic image data be reproduced as reproduces dynamic image data for it, it is characterized in that, this dynamic image decoding device has: decoding unit, and its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded; The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated the decoder object piece by the coded data after the decoding unit decodes; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; Retrieval unit, it is from the dynamic image data of reproduction by cell stores, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal with the position relation of regulation and decoder object piece in abutting connection with and belong to the dynamic image data of reproduction by cell stores; And prediction signal decision unit, it is based on the position relation of pixel groups that is retrieved by retrieval unit and regulation, according to promptly predicting piece by the prediction signal of reproducing dynamic image data decision decoder object piece of cell stores; The difference signal that decoding unit generates the decoder object piece is the decoded difference value piece; The reproduced image generation unit is that unit is with decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.According to this structure, can be to being correctly decoded by the dynamic image data behind the above-mentioned moving picture encoding device coding.
Preferably, dynamic image decoding device also has: estimation unit, and it compares with template with by the dynamic image data of reproduction of cell stores, estimates the spatial continuity of the image of decoder object piece according to this comparative result; And setup unit, it further cuts apart the decoder object piece according to the spatial continuity of the image that is estimated by estimation unit, and the decoder object piece after this is cut apart is set at new decoder object piece, and sets the template at this new decoder object piece.According to this structure, can be to being correctly decoded by the dynamic image data behind the above-mentioned moving picture encoding device coding.
Preferably, coding unit makes the processing of dwindling that pixel count reduces by utilizing prescriptive procedure, difference block is generated as pixel count dwindles difference block than what this difference block was lacked, and this is dwindled difference block encode; The reproduced image generation unit generates the reproducing signal dwindle difference block and promptly decodes and dwindle difference block, by the processing and amplifying of utilizing prescriptive procedure that pixel count is increased, dwindles difference block generating solution code block according to this decoding.According to this structure owing to can reduce the pixel count that dwindles difference block of coded object, thereby for estimated performance with respect to regional low flat site with strong feature, do not reduce the sign indicating number that image quality can efficient reduces prediction signal well and measure.
Preferably, decoding unit is by decoding to coded data, and difference block is dwindled in the decoding that the generation pixel count lacks than difference block; The processing and amplifying of reproduced image generation unit by utilizing prescriptive procedure that pixel count is increased dwindled difference block generating solution code block according to decoding.According to this structure, can be to being correctly decoded by the dynamic image data behind the above-mentioned moving picture encoding device coding.
Moving picture encoding device of the present invention, it is that unit encodes to dynamic image data with the piece, it is characterized in that this moving picture encoding device has: cutting unit, its two field picture that will constitute dynamic image data are divided into a plurality of coded object pieces as the zone that becomes coded object; Coding unit, it is encoded to the coded object piece; The reproduced image generation unit, its reproducing signal that generates the coded object piece is a decoding block; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generation unit, its according to the position relation of regulation and coded object piece in abutting connection with and belong to the reproducing signal that reproduces dynamic image data by cell stores, the prediction signal of using predetermined method to generate the coded object piece is promptly predicted piece; Coding unit is a difference block at the difference signal that to be unit with the pixel deduct this coded object piece behind the prediction piece from the coded object piece, by utilizing prescriptive procedure to make the processing of dwindling of pixel count minimizing, generate pixel count and dwindle difference block, and this is dwindled difference block encode than what this difference block was lacked; The reproduced image generation unit generates the reproducing signal dwindle difference block and promptly decodes and dwindle difference block, by the processing and amplifying of utilizing prescriptive procedure that pixel count is increased, dwindles difference block generating solution code block according to this decoding.
In moving picture encoding device of the present invention, according to concerning with the position of stipulating with coded object piece adjacency and belong to the reproducing signal that reproduces dynamic image data that the prediction signal that generates the coded object piece is promptly predicted piece.And, generate the difference block of dwindling of coded object that pixel count lacks than above-mentioned difference block according to the prediction piece.Promptly, according to moving picture encoding device of the present invention, owing to can reduce the pixel count that dwindles difference block of coded object, thus for estimated performance with respect to regional low flat site with strong feature, do not reduce the sign indicating number that image quality can efficient reduces prediction signal well and measure.
Dynamic image decoding device of the present invention, to be unit with the piece with the coded data of dynamic image data be reproduced as reproduces dynamic image data for it, it is characterized in that, this dynamic image decoding device has: decoding unit, and its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded; The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated the decoder object piece by the coded data after the decoding unit decodes; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generation unit, its according to the position relation of regulation and decoder object piece in abutting connection with and belong to the reproducing signal that reproduces dynamic image data by cell stores, the prediction signal of using predetermined method to generate the decoder object piece is promptly predicted piece; Decoding unit is by decoding to coded data, and the difference signal that generates pixel count ratio decoder object piece is that difference block is dwindled in the few decoding of difference block; The processing and amplifying of reproduced image generation unit by utilizing prescriptive procedure that pixel count is increased dwindled difference block generating solution code block according to decoding.According to this structure, can be to being correctly decoded by the dynamic image data behind the above-mentioned moving picture encoding device coding.
Preferably, coding unit is applied to coded object piece and prediction piece and obtains dwindling piece respectively and dwindle the prediction piece dwindling to handle, and to be unit with the pixel dwindle piece from this deducts this and dwindle the prediction piece and generate and dwindle difference block; The reproducing signal that dwindles difference block after the generation of reproduced image generation unit is encoded by coding unit is promptly decoded and is dwindled difference block, be that unit dwindles difference block with this decoding and dwindles the addition of prediction piece and generates decoding and dwindle piece with the pixel, and processing and amplifying is applied to this decoding dwindles piece and generate decoding block.According to this structure, owing to can generate the difference block of dwindling of coded object reliably, thereby can implement the present invention reliably.
Preferably, coding unit is by be applied to difference block and generate and dwindle difference block dwindling to handle; The reproducing signal that dwindles difference block after the generation of reproduced image generation unit is encoded by coding unit is promptly decoded and is dwindled piece, dwindle piece and generate the decoded difference value piece by processing and amplifying being applied to this decoding, and be that unit generates decoding block with this decoded difference value piece and the addition of prediction piece with the pixel.According to this structure, owing to can generate the difference block of dwindling of coded object reliably, thereby can implement the present invention reliably.
Preferably, the reproduced image generation unit will dwindle to handle and is applied to predict piece and obtains dwindling the prediction piece, to be unit with the pixel dwindle difference block and this with decoding dwindles the addition of prediction piece and generates decoding and dwindle piece, and processing and amplifying is applied to this decoding dwindles piece and generate decoding block.According to this structure, can be to being correctly decoded by the dynamic image data behind the above-mentioned moving picture encoding device coding.
Preferably, the reproduced image generation unit dwindles difference block and generates the decoded difference value piece by processing and amplifying being applied to decoding, and is that unit generates decoding block with this decoded difference value piece and the addition of prediction piece with the pixel.According to this structure, can be to being correctly decoded by the dynamic image data behind the above-mentioned moving picture encoding device coding.
Preferably, 1 template is selected in prediction signal decision unit from variform a plurality of templates.According to this structure, the generation forecast piece can improve encoding process efficient effectively.
Preferably, 1 template is selected with reference to by the reproducing signal that reproduces dynamic image data of cell stores or the information relevant with this reproducing signal in prediction signal decision unit.According to this structure, can suitably select template.
Preferably, coding unit is encoded to the information of specifying the template of being selected by prediction signal decision unit.According to this structure, make the template in the dynamic image decoding device select to become easily, can carry out more effective decoding.
Preferably, 1 template is selected in prediction signal decision unit from variform a plurality of templates.According to this structure, can be to being correctly decoded by the dynamic image data behind the above-mentioned moving picture encoding device coding.
Preferably, 1 template is selected with reference to by the reproducing signal that reproduces dynamic image data of cell stores or the information relevant with this reproducing signal in prediction signal decision unit.According to this structure, can be to being correctly decoded by the dynamic image data behind the above-mentioned moving picture encoding device coding.
Preferably, decoding unit is decoded to the information of specifying selected template; 1 template is selected with reference to the information by the selected template of appointment after the decoding unit decodes in prediction signal decision unit from variform a plurality of templates.According to this structure, can be to being correctly decoded by the dynamic image data behind the above-mentioned moving picture encoding device coding.
In addition, the present invention also can be described except the invention that can be used as moving picture encoding device and dynamic image decoding device as mentioned above is described in the invention as motion image encoding method, dynamic image encoding, dynamic image decoding method and moving image decoding program as described below.This is a difference such as classification and the identical invention of essence obtains identical effect and effect only.
Motion image encoding method of the present invention, motion image encoding method in it is the moving picture encoding device that to be unit with the piece encode to dynamic image data, it is characterized in that, this motion image encoding method has: segmentation procedure, and its two field picture that will constitute dynamic image data is divided into a plurality of coded object pieces; Coding step, it is encoded to the coded object piece; Reproduced image generates step, and its reproducing signal that generates the coded object piece is a decoding block; Storing step, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generates step, its use according to the position relation of regulation and coded object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data stored generates in storing step, the prediction signal that generates the coded object piece is promptly predicted piece; In coding step, to be unit with the pixel deduct the difference signal that the prediction piece generates this coded object piece from the coded object piece is difference block, and this difference block is encoded; In reproduced image generated step, the reproducing signal that is created on the difference block behind the coding in the coding step was the decoded difference value piece, and was that unit is with this decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.
Motion image encoding method of the present invention, motion image encoding method in it is the moving picture encoding device that to be unit with the piece encode to dynamic image data, it is characterized in that, this motion image encoding method has: segmentation procedure, and its two field picture that will constitute dynamic image data is divided into a plurality of coded object pieces; Coding step, it is encoded to the coded object piece; Reproduced image generates step, and its reproducing signal that generates the coded object piece is a decoding block; Storing step, the dynamic image data of reproduction that its storage generates according to reproducing signal; Searching step, it is from the dynamic image data of storing storing step of reproduction, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal concerns with coded object piece adjacency with the position of stipulating and belong to the dynamic image data of storing of reproduction in storing step; And the prediction signal deciding step, it promptly predicts piece based on the position relation of pixel groups that retrieves and regulation according to the prediction signal of reproducing dynamic image data decision coded object piece of storing in storing step in searching step; In coding step, to be unit with the pixel deduct the difference signal that the prediction piece generates this coded object piece from the coded object piece is difference block, and this difference block is encoded; In reproduced image generated step, the reproducing signal that is created on the difference block behind the coding in the coding step was the decoded difference value piece, and was that unit is with this decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.
Dynamic image decoding method of the present invention, it is, and to be unit with the piece be reproduced as dynamic image decoding method in the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data, it is characterized in that, this dynamic image decoding method has: decoding step, and its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded; Reproduced image generates step, and it is a decoding block according to the reproducing signal that decoded coded data in decoding step generates the decoder object piece; Storing step, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generates step, its use according to the position relation of regulation and decoder object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data stored generates in storing step, the prediction signal that generates this decoder object piece is promptly predicted piece; In decoding step, the difference signal that generates the decoder object piece is the decoded difference value piece; Generating in the step at reproduced image, is that unit is with decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.
Dynamic image decoding method of the present invention, it is, and to be unit with the piece be reproduced as dynamic image decoding method in the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data, it is characterized in that, this dynamic image decoding method has: decoding step, and its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded; Reproduced image generates step, and it is a decoding block according to the reproducing signal that decoded coded data in decoding step generates the decoder object piece; Storing step, the dynamic image data of reproduction that its storage generates according to reproducing signal; Searching step, it is from the dynamic image data of storing storing step of reproduction, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal concerns with decoder object piece adjacency with the position of stipulating and belong to the dynamic image data of storing of reproduction in storing step; And the prediction signal deciding step, it promptly predicts piece based on the position relation of pixel groups that retrieves and regulation according to the prediction signal of reproducing dynamic image data decision decoder object piece of storing in storing step in searching step; In decoding step, the difference signal that generates the decoder object piece is the decoded difference value piece; Generating in the step at reproduced image, is that unit is with decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.
Motion image encoding method of the present invention, motion image encoding method in it is the moving picture encoding device that to be unit with the piece encode to dynamic image data, it is characterized in that, this motion image encoding method has: segmentation procedure, its two field picture that will constitute dynamic image data are divided into a plurality of coded object pieces as the zone that becomes coded object; Coding step, it is encoded to the coded object piece; Reproduced image generates step, and its reproducing signal that generates the coded object piece is a decoding block; Storing step, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generates step, it concerns with coded object piece adjacency according to the position with regulation and belongs to the reproducing signal of storing that reproduces dynamic image data that in storing step the prediction signal of using predetermined method to generate the coded object piece is promptly predicted piece; In coding step, at the difference signal that to be unit with the pixel deduct this coded object piece behind the prediction piece from the coded object piece is difference block, by utilizing prescriptive procedure to make the processing of dwindling of pixel count minimizing, generate pixel count and dwindle difference block, and this is dwindled difference block encode than what this difference block was lacked; In reproduced image generates step, generate the reproducing signal that dwindles difference block and promptly decode and dwindle difference block, and the processing and amplifying by utilizing prescriptive procedure that pixel count is increased, dwindle difference block generating solution code block according to this decoding.
Dynamic image decoding method of the present invention, it is, and to be unit with the piece be reproduced as dynamic image decoding method in the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data, it is characterized in that, this dynamic image decoding method has: decoding step, and its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded; Reproduced image generates step, and it is a decoding block according to the reproducing signal that decoded coded data in decoding step generates the decoder object piece; Storing step, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generates step, it concerns with decoder object piece adjacency according to the position with regulation and belongs to the reproducing signal of storing that reproduces dynamic image data that in storing step the prediction signal of using predetermined method to generate the decoder object piece is promptly predicted piece; In decoding step, by coded data is decoded, the difference signal that generates pixel count ratio decoder object piece is that difference block is dwindled in the few decoding of difference block; Generate in the step at reproduced image,, dwindle difference block generating solution code block according to decoding by the processing and amplifying of utilizing prescriptive procedure that pixel count is increased.
Dynamic image encoding of the present invention, its control is the moving picture encoding device that unit encodes to dynamic image data with the piece, it is characterized in that, this dynamic image encoding makes moving picture encoding device as playing a role with lower unit: cutting unit, and its two field picture that will constitute dynamic image data is divided into a plurality of coded object pieces; Coding unit, it is encoded to the coded object piece; The reproduced image generation unit, its reproducing signal that generates the coded object piece is a decoding block; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generation unit, its use according to the position relation of regulation and coded object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data by cell stores generates, the prediction signal that generates the coded object piece is promptly predicted piece; To be unit with the pixel deduct the difference signal that the prediction piece generates this coded object piece from the coded object piece to coding unit is difference block, and this difference block is encoded; The reproducing signal that the reproduced image generation unit generates by the difference block behind the coding unit coding is the decoded difference value piece, and is that unit is with this decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.
Dynamic image encoding of the present invention, its control is the moving picture encoding device that unit encodes to dynamic image data with the piece, it is characterized in that, this dynamic image encoding makes moving picture encoding device as playing a role with lower unit: cutting unit, and its two field picture that will constitute dynamic image data is divided into a plurality of coded object pieces; Coding unit, it is encoded to the coded object piece; The reproduced image generation unit, its reproducing signal that generates the coded object piece is a decoding block; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; Retrieval unit, it is from the dynamic image data of reproduction by cell stores, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal with the position relation of regulation and coded object piece in abutting connection with and belong to the dynamic image data of reproduction by cell stores; And prediction signal decision unit, it is based on the position relation of pixel groups that is retrieved by retrieval unit and regulation, according to promptly predicting piece by the prediction signal of reproducing dynamic image data decision coded object piece of cell stores; To be unit with the pixel deduct the difference signal that the prediction piece generates this coded object piece from the coded object piece to coding unit is difference block, and this difference block is encoded; The reproducing signal that the reproduced image generation unit generates by the difference block behind the coding unit coding is the decoded difference value piece, and is that unit is with this decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.
Moving image decoding program of the present invention, to be unit with the piece be reproduced as the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data in its control, it is characterized in that, this moving image decoding program makes dynamic image decoding device as playing a role with lower unit: decoding unit, and its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded; The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated the decoder object piece by the coded data after the decoding unit decodes; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generation unit, its use according to the position relation of regulation and decoder object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data by cell stores generates, the prediction signal that generates this decoder object piece is promptly predicted piece; The difference signal that decoding unit generates the decoder object piece is the decoded difference value piece; The reproduced image generation unit is that unit is with decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.
Moving image decoding program of the present invention, to be unit with the piece be reproduced as the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data in its control, it is characterized in that, this moving image decoding program makes dynamic image decoding device as playing a role with lower unit: decoding unit, and its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded; The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated the decoder object piece by the coded data after the decoding unit decodes; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; Retrieval unit, it is from the dynamic image data of reproduction by cell stores, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal with the position relation of regulation and decoder object piece in abutting connection with and belong to the dynamic image data of reproduction by cell stores; And prediction signal decision unit, it is based on the position relation of pixel groups that is retrieved by retrieval unit and regulation, according to promptly predicting piece by the prediction signal of reproducing dynamic image data decision decoder object piece of cell stores; The difference signal that decoding unit generates the decoder object piece is the decoded difference value piece; The reproduced image generation unit is that unit is with decoded difference value piece and the addition of prediction piece, generating solution code block with the pixel.
Dynamic image encoding of the present invention, its control is the moving picture encoding device that unit encodes to dynamic image data with the piece, it is characterized in that, this dynamic image encoding makes moving picture encoding device as playing a role with lower unit: cutting unit, its two field picture that will constitute dynamic image data are divided into a plurality of coded object pieces as the zone that becomes coded object; Coding unit, it is encoded to the coded object piece; The reproduced image generation unit, its reproducing signal that generates the coded object piece is a decoding block; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generation unit, its according to the position relation of regulation and coded object piece in abutting connection with and belong to the reproducing signal that reproduces dynamic image data by cell stores, the prediction signal of using predetermined method to generate the coded object piece is promptly predicted piece; Coding unit is a difference block at the difference signal that to be unit with the pixel deduct this coded object piece behind the prediction piece from the coded object piece, by utilizing prescriptive procedure to make the processing of dwindling of pixel count minimizing, generate pixel count and dwindle difference block, and this is dwindled difference block encode than what this difference block was lacked; The reproduced image generation unit generates the reproducing signal dwindle difference block and promptly decodes and dwindle difference block, and the processing and amplifying by utilizing prescriptive procedure that pixel count is increased, and dwindles difference block generating solution code block according to this decoding.
Moving image decoding program of the present invention, to be unit with the piece be reproduced as the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data in its control, it is characterized in that, this moving image decoding program makes dynamic image decoding device as playing a role with lower unit: decoding unit, and its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded; The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated the decoder object piece by the coded data after the decoding unit decodes; Memory cell, the dynamic image data of reproduction that its storage generates according to reproducing signal; And prediction signal generation unit, its according to the position relation of regulation and decoder object piece in abutting connection with and belong to the reproducing signal that reproduces dynamic image data by cell stores, the prediction signal of using predetermined method to generate the decoder object piece is promptly predicted piece; Decoding unit is by decoding to coded data, and the difference signal that generates pixel count ratio decoder object piece is that difference block is dwindled in the few decoding of difference block; The processing and amplifying of reproduced image generation unit by utilizing prescriptive procedure that pixel count is increased dwindled difference block generating solution code block according to decoding.
Moving picture encoding device of the present invention, it is encoded to dynamic image data, it is characterized in that this moving picture encoding device has: cutting unit, its two field picture that will constitute dynamic image data are divided into a plurality of zones as the zone that becomes coded object; Coding unit, it is encoded to each the regional image after being cut apart by cutting unit; The reproduced image generation unit, it generates the reproduced image by the image behind the coding unit coding; Memory cell, the reproduced image that its storage is generated by the reproduced image generation unit; Retrieval unit, its from reproduced image by cell stores, retrieval with the zone of the position relation of regulation and the image of the coded object that becomes coding unit in abutting connection with and with as zone by the high image of the reproduced image correlation in the template zone of the part of the reproduced image of cell stores; And prediction signal decision unit, it is based on position relation of zone that is retrieved by retrieval unit and regulation, according to the prediction signal by the zone of reproduced image decision the becoming coded object of cell stores; Coding unit generates by the prediction signal of prediction signal decision unit decision and becomes difference signal between the image in zone of coded object, and this difference signal is encoded.
In moving picture encoding device of the present invention, at first, from reproduced image retrieval with the zone of the position relation of regulation and the image that becomes coded object in abutting connection with and with the zone of the high image of the reproduced image correlation in template zone.Next, based on the position of zone that retrieves and afore mentioned rules relation, according to the prediction signal in the zone of reproduced image decision becoming coded object.Use this prediction signal to encode.That is,, do not use motion vector can determine prediction signal, can carry out efficient coding according to moving picture encoding device of the present invention.
Dynamic image decoding device of the present invention, dynamic image data after it is encoded to the two field picture that is divided into a plurality of zones is decoded, it is characterized in that this dynamic image decoding device has: decoding unit, it is decoded to each the regional data after encoding; The reproduced image generation unit, it is according to generating reproduced image by the image after the decoding unit decodes; Memory cell, the image that its storage is generated by the reproduced image generation unit; Retrieval unit, its from reproduced image by cell stores, retrieval with the zone of the position relation of regulation and the image of the decoder object that becomes decoding unit in abutting connection with and with as zone by the high image of the reproduced image correlation in the template zone of the part of the reproduced image of cell stores; And prediction signal decision unit, it is based on position relation of zone that is retrieved by retrieval unit and regulation, according to the prediction signal by the zone of reproduced image decision the becoming decoder object of cell stores; The reproduced image generation unit is by generating by the prediction signal of prediction signal decision unit decision with by obtaining reproduced image with signal between the image after the decoding unit decodes.
In dynamic image decoding device of the present invention, can determine prediction signal in the same manner with above-mentioned moving picture encoding device, dynamic image is decoded.That is,, can be correctly decoded carried out the dynamic image data after the efficient coding by above-mentioned moving picture encoding device according to dynamic image decoding device of the present invention.
And, preferably, moving picture encoding device also has: estimation unit, and it compares with the reproduced image in template zone with by the reproduced image of cell stores, and estimates to become the spatial continuity of image in the zone of coded object according to this comparative result; And setup unit, it is according to the spatial continuity of the image that is estimated by estimation unit, further be partitioned into the zone of coded object, this divided area is set at new coded object zone, and set template zone at this new coded object zone.According to this structure owing to can suitably select the area size of prediction signal according to the spatial continuity of reproduced image, even thereby change in the coding of violent dynamic image data at amount of exercise, also can improve code efficiency.
And, preferably, dynamic image decoding device also has: estimation unit, and it compares with the reproduced image in template zone with by the reproduced image of cell stores, estimates to become the spatial continuity of image in the zone of decoder object according to this comparative result; And setup unit, it is according to the spatial continuity of the image that is estimated by estimation unit, further be partitioned into the zone of decoder object, this divided area is set at new decoder object zone, and set template zone at this new decoder object zone.According to this structure, can be to being correctly decoded by the dynamic image data behind the above-mentioned moving picture encoding device coding.
In addition, the present invention also can be described except the invention that can be used as moving picture encoding device and dynamic image decoding device as mentioned above is described in the invention as motion image encoding method, dynamic image encoding, dynamic image decoding method and moving image decoding program as described below.This is a difference such as classification and the identical invention of essence obtains identical effect and effect only.
Motion image encoding method of the present invention, it is the motion image encoding method in the moving picture encoding device that dynamic image data is encoded, it is characterized in that, this motion image encoding method has: segmentation procedure, its two field picture that will constitute dynamic image data are divided into a plurality of zones as the zone that becomes coded object; Coding step, it is encoded to each the regional image after cutting apart in segmentation procedure; Reproduced image generates step, and it is created on the reproduced image of the image behind the coding in the coding step; Storing step, it is stored in reproduced image and generates the reproduced image that generates in the step; Searching step, its from the reproduced image of storing step, storing, retrieval with the zone of the position relation of regulation and the image that in coding step, becomes coded object in abutting connection with and with zone as the high image of the reproduced image correlation in the template zone of the part of the reproduced image of in storing step, storing; And the prediction signal deciding step, it is based on position of zone that retrieves in searching step and regulation relation, according to the prediction signal in the zone of reproduced image decision the becoming coded object of storing in storing step; In coding step, be created on the prediction signal that determines in the prediction signal deciding step and become difference signal between the image in zone of coded object, and this difference signal is encoded.
Dynamic image encoding of the present invention, it is a dynamic image encoding of controlling the moving picture encoding device that dynamic image data is encoded, it is characterized in that, this dynamic image encoding makes moving picture encoding device as playing a role with lower unit: cutting unit, its two field picture that will constitute dynamic image data are divided into a plurality of zones as the zone that becomes coded object; Coding unit, it is encoded to each the regional image after being cut apart by cutting unit; The reproduced image generation unit, it generates the reproduced image by the image behind the coding unit coding; Memory cell, the reproduced image that its storage is generated by the reproduced image generation unit; Retrieval unit, its from reproduced image by cell stores, retrieval with the zone of the position relation of regulation and the image of the coded object that becomes coding unit in abutting connection with and with as zone by the high image of the reproduced image correlation in the template zone of the part of the reproduced image of cell stores; And prediction signal decision unit, it is based on position relation of zone that is retrieved by retrieval unit and regulation, according to the prediction signal by the zone of reproduced image decision the becoming coded object of cell stores; Coding unit generates by the prediction signal of prediction signal decision unit decision and becomes difference signal between the image in zone of coded object, and this difference signal is encoded.
Dynamic image decoding method of the present invention, it is the dynamic image decoding method in the dynamic image decoding device of decoding of the dynamic image data after the two field picture that is divided into a plurality of zones is encoded, it is characterized in that, this dynamic image decoding method has: decoding step, and it is decoded to each the regional data after encoding; Reproduced image generates step, and it generates reproduced image according to decoded image in decoding step; Storing step, it is stored in reproduced image and generates the image that generates in the step; Searching step, its from the reproduced image of storing step, storing, retrieval with the zone of the position relation of regulation and the image of the decoder object that becomes decoding step in abutting connection with and with zone as the high image of the reproduced image correlation in the template zone of the part of the reproduced image of in storing step, storing; And the prediction signal deciding step, it is based on position of zone that retrieves in searching step and regulation relation, according to the prediction signal in the zone of reproduced image decision the becoming decoder object of storing in storing step; In reproduced image generates step, by being created on the prediction signal that determines in the prediction signal deciding step and in decoding step, obtaining reproduced image with signal between the decoded image.
Moving image decoding program of the present invention, it is the moving image decoding program of the dynamic image decoding device of decoding of the dynamic image data after control is encoded to the two field picture that is divided into a plurality of zones, it is characterized in that, this moving image decoding program makes dynamic image decoding device as playing a role with lower unit: decoding unit, and it is decoded to each the regional data after encoding; The reproduced image generation unit, it is according to generating reproduced image by the image after the decoding unit decodes; Memory cell, the image that its storage is generated by the reproduced image generation unit; Retrieval unit, its from reproduced image by cell stores, retrieval with the zone of the position relation of regulation and the image of the decoder object that becomes decoding unit in abutting connection with and with as zone by the high image of the reproduced image correlation in the template zone of the part of the reproduced image of cell stores; And prediction signal decision unit, it is based on position relation of zone that is retrieved by retrieval unit and regulation, according to the prediction signal by the zone of reproduced image decision the becoming decoder object of cell stores; The reproduced image generation unit is by generating by the prediction signal of prediction signal decision unit decision with by obtaining reproduced image with signal between the image after the decoding unit decodes.
According to the present invention, the zone adjacency and the reproduction regions high that concern and become the image of coded object owing to retrieval with the position of regulation with the template area coherence, and decide prediction signal, thereby do not use motion vector can carry out efficient coding according to this zone that retrieves and above-mentioned position relation.
Description of drawings
Fig. 1 is the figure of structure that the moving picture encoding device of the 1st execution mode of the present invention is shown.
Fig. 2 is the figure that the structure of the prediction generating unit in the moving picture encoding device is shown.
Fig. 3 is the figure that the position relation in template zone and forecasting object zone is shown.
Fig. 4 is used for using template matches to determine the figure that the detailed action of prediction signal describes.
Fig. 5 is the flow chart that the processing of being carried out by the moving picture encoding device of the 1st execution mode of the present invention is shown.
Fig. 6 is the figure of structure that the dynamic image decoding device of the 1st execution mode of the present invention is shown.
Fig. 7 is the flow chart that the processing of being carried out by the dynamic image decoding device of the 1st execution mode of the present invention is shown.
Fig. 8 is the figure that the structure of the prediction generating unit in the 2nd execution mode is shown.
Fig. 9 is the figure that is illustrated in the forecasting object zone of being cut apart in the 2nd execution mode.
Figure 10 is the flow chart that the processing of being carried out by the moving picture encoding device of the 2nd execution mode of the present invention is shown.
Figure 11 is the flow chart that the processing of being carried out by the dynamic image decoding device of the 2nd execution mode of the present invention is shown.
Figure 12 is the figure that coded sequence is shown.
Figure 13 is the figure that the position relation example in template zone corresponding with coded sequence and forecasting object zone is shown.
Figure 14 is used for the figure that the function of judging part described at the 2nd execution mode.
Figure 15 is used for the figure that the function of judging part described at the 2nd execution mode.
Figure 16 is the figure of structure that the dynamic image encoding of embodiments of the present invention is shown.
Figure 17 is the figure of structure that the moving image decoding program of embodiments of the present invention is shown.
Figure 18 is the figure of structure of variation that the moving picture encoding device of the 1st execution mode is shown.
Figure 19 is the figure of structure of variation that the dynamic image decoding device of the 1st execution mode is shown.
Figure 20 is the figure of structure of variation that the prediction generating unit of the 1st execution mode is shown.
Figure 21 is the figure of structure that the moving picture encoding device of the 3rd execution mode of the present invention is shown.
Figure 22 is the figure of structure that the dynamic image decoding device of the 3rd execution mode of the present invention is shown.
Figure 23 is the figure that the dwindling of piece in the 3rd execution mode/processing and amplifying is shown.
Figure 24 is the flow chart that the processing of being carried out by the moving picture encoding device of the 3rd execution mode of the present invention is shown.
Figure 25 is the flow chart that the processing of being carried out by the dynamic image decoding device of the 3rd execution mode of the present invention is shown.
Figure 26 is the figure of structure of variation that the moving picture encoding device of the 3rd execution mode is shown.
Figure 27 is the figure of structure of variation that the dynamic image decoding device of the 3rd execution mode is shown.
Figure 28 is the figure that another example of the dwindling of piece in the 3rd execution mode/processing and amplifying is shown.
Figure 29 is the figure that another example of the dwindling of piece in the 3rd execution mode/processing and amplifying is shown.
Figure 30 is the figure of structure of variation that the moving picture encoding device of the 3rd execution mode is shown.
Figure 31 is the figure of structure of variation that the dynamic image decoding device of the 3rd execution mode is shown.
Figure 32 is the figure of structure of variation that the moving picture encoding device of the 3rd execution mode is shown.
Figure 33 is the figure of structure of variation that the dynamic image decoding device of the 3rd execution mode is shown.
Figure 34 is the figure of structure of variation that the moving picture encoding device of the 3rd execution mode is shown.
Figure 35 is the figure of structure of variation that the dynamic image decoding device of the 3rd execution mode is shown.
Figure 36 is the figure of structure of variation that the prediction generating unit of the 3rd execution mode is shown.
Figure 37 is the figure that the prediction processing example in the 3rd execution mode is shown.
Symbol description
100,1800,2100,2600,3000,3200,3400: moving picture encoding device; 600,1900,2200,2700,3100,3300,3500: dynamic image decoding device; 101: Region Segmentation portion; 102,2102: subtraction portion; 103,2103: transformation component; 104,2104,3004: encoding section; 105,602,2105,2202: inverse transformation portion; 106,603,2106,2203: addition portion; 107,604: storage part; 108,605,800,1108,3408,3505,3608: the prediction generating unit; 201,802,3602: template zone determination section 202: matching part 203,3603: fill up portion; 601,2201,3101: lsb decoder; 801: judging part; 109,2612,3012: selection portion; 204: the signal generating unit; 2110,2207,3210: dwindle portion; 2111,2208,3211,3308: the enlarging section; 2613,2709,3109: switching part; 3013,3110: switch 1600,1700: recording medium 1600a, 1700a: program storage area 1601: dynamic image encoding; 1701: the moving image decoding program; 1601a, 1701a: primary module; 1601b: Region Segmentation module; 1601c: subtraction block; 1601d: conversion module 1601e: coding module; 1601f, 1701c: inverse transform module; 1601g, 1701d: addition module; 1601h, 1701e: memory module; 1601i, 1701f: prediction generation module; 1601j, 1701g: template zone decision module; 1601k, 1701h: matching module; 1601m, 1701i: fill up module; 1701b: decoder module.
Embodiment
Below, the preferred implementation to moving picture encoding device of the present invention, motion image encoding method, dynamic image encoding, dynamic image decoding device, dynamic image decoding method and moving image decoding program is elaborated in conjunction with the accompanying drawings.In addition, in description of drawings, identical element is enclosed same-sign, omit repeat specification.
[the 1st execution mode]
Fig. 1 illustrates the moving picture encoding device 100 of the 1st execution mode.Moving picture encoding device 100 is to be that unit carries out apparatus for encoding to dynamic image data with the piece.Moving picture encoding device 100 comes dynamic image data is encoded by the two field picture that constitutes dynamic image data is encoded to this two field picture in turn as input.CPU), the hardware such as information processor of frame memory, hard disk etc. realize moving picture encoding device 100 is by having CPU (Central Processing Unit:.Moving picture encoding device 100 moves by making above-mentioned hardware inscape, realizes the function inscape of following explanation.
As shown in Figure 1, moving picture encoding device 100 constitutes and has: Region Segmentation portion 101, subtraction portion 102, transformation component 103, encoding section 104, inverse transformation portion 105, addition portion 106, storage part 107 and prediction generating unit 108.Below, the function of each one is described.
Template zone determination section 201 is to decide a template zone and an inscape this regional signal (template), retrieval unit that is used for the generation forecast signal according to the input from storage part 107.That is, template zone determination section 201 be according to the position relation of regulation and coded object piece in abutting connection with and the reproducing signal that reproduces dynamic image data that belongs to by storage part 17 storages generate inscape template, retrieval unit.And the prediction signal that template zone determination section 201 also is to use template to generate the coded object piece is promptly predicted an inscape of the prediction signal generation unit of piece.As shown in Figure 3, template zone 301 be with the zone 302 of the position relation of regulation and the image that becomes coded object in abutting connection with and by the zone of the reproduced image of storage part 107 storages, the pixel region that promptly constitutes by the reproducing signal that reproduces dynamic image data.Specifically, template zone is corresponding is and is stored in the reproduction regions of the forecasting object piece same number of frames in the storage part 107, promptly comprises to be present in and the pixel region of the forecasting object block space adjoining position pixel groups in interior prescribed level.Therefore, the coded sequence (exporting and carry out the order of encoding process from Region Segmentation portion 101) of piece is depended in the position in template zone.Storage in advance is used to determine the condition of the following explanation in template zone in the determination section 201 of template zone.
Figure 13 illustrates an example of the position relation in template zone and forecasting object zone.Shown in Figure 13 (a), when coded sequence was pressed raster scan order and carried out, template zone 1301 became the left side that is positioned at forecasting object piece 1302 and the zone of upside.Shown in Figure 13 (a), being positioned at the left side of forecasting object piece 1302 and the zone 1303 of upside is the zone of preserving reproduced image at storage part 107.And, when coded sequence is the zigzag order, can be according to the position that degree changes the template zone of carrying out of coding.When shown in Figure 13 (b) from the picture left side when encode in the right, forecasting object piece 1304 becomes the right side that is positioned at forecasting object piece 1305 and the zone of upside.When shown in Figure 13 (c) from picture the right when encode in the left side, forecasting object piece 1304 becomes the left side that is positioned at forecasting object piece 1305 and the zone of upside.
Matching part 202 is to use the reproducing signal by the template zone of template zone determination section 201 decisions, the reproduced image of being preserved by storage part 107 is implemented template matches as search domain, and the retrieval unit in the zone that the correlation of the reproducing signal in retrieval (detections) and template zone is the highest in search domain.That is, matching part 202 is the retrieval units from the high pixel groups of the template correlation of being retrieved the dynamic image data by reproducing of storage part 107 storages and determined by template zone determination section 201.And the prediction signal that matching part 202 also is to use template to generate the coded object piece is promptly predicted an inscape of the prediction signal generation unit of piece.The back is described in more detail template matches.
Fill up portion 203 and be based on the zone (high correlation zone) that retrieves by matching part 202 and the relation of the position between forecasting object piece and the template zone, determine the unit according to the prediction signal of the prediction signal of reproduced image setting and decision and forecasting object piece same size.That is, fill up portion 203 and be based on pixel groups and the above-mentioned position relation that retrieves by matching part 202, according to the prediction signal decision unit of promptly predicting piece by the prediction signal of reproducing dynamic image data decision coded object piece of storage part 107 storages.And, fill up the inscape that prediction signal that portion 203 also is to use template to generate the coded object piece is promptly predicted the prediction signal generation unit of piece.The high correlation that retrieves zone and concern identical with position between template zone and the forecasting object piece as the position relation between the zone of the image of prediction signal.For example, when the coded sequence of piece is raster scan order, become the zone of prediction signal with the zone of the right side in high correlation zone and downside adjacency.The prediction signal that is determined is imported into subtraction portion 102 and addition portion 106 as the output from prediction generating unit 108.
Use Fig. 4 that matching part 202 and the detailed action of filling up portion's 203 use template matches decision prediction signal are described.The template matches at position like the images category of retrieval and template zone 401 in the range of search 403,404 is carried out in matching part 202.Range of search includes: comprise the image 404 that reproduces pixel region 403 and the frame that has reproduced in addition in the frame (forecasting object frame) in template zone 401 and forecasting object zone 402.At first, measure the signal in template zone 401 and be positioned at any place and the correlation between signals identical shaped pixel groups of template regional 401 of range of search.As the desired value of the correlation of expression this moment, can use the expression difference signal absolute value and the average MSE (mean square error) of the square error of SAD (sum of absolute difference), expression difference signal etc.Obtain the desired value of correlation at all pixel groups that in range of search, can obtain, the information (address) of the pixel groups (but, except the image in template zone 401 self) of expression desired value minimum is exported as result for retrieval.
Fill up portion 203 and will be set at prediction signal with the high correlation zone pixel adjacent group that retrieves by template matches.In the place high, because the possibility height that correlation uprises, thereby also can be used as prediction signal with the zone of template zone adjacency with the template area coherence.Therefore, use the Forecasting Methodology of this method to set up.
Next, use the flow chart of Fig. 5 that the encoding process (motion image encoding method) in the moving picture encoding device 100 of present embodiment is described.
At first, when the dynamic image data of coded object is imported into moving picture encoding device 100, be imported into Region Segmentation portion 101 according to the every two field picture that constitutes dynamic image.The two field picture of being imported is divided into a plurality of (S501, segmentation procedure) of prescribed level by Region Segmentation portion 101.Later processing all is that unit carries out with the piece.Piece is imported into subtraction portion 102 as the image in coded object zone.
Then, the prediction signal that generates the coded object piece by prediction generating unit 108 as described below.In addition, prediction generating unit 108 is by the coded sequence of controller (not illustrating) the notice piece of blanket encoding process in moving picture encoding device 100.Perhaps, predict generating unit 108 coded sequence of memory block in advance.At first, by template zone determination section 201 with the reproduced image of piece adjacency on decision template zone (S502, searching step).Then, carry out template matches, the high zone (S503, searching step) of reproducing signal correlation in retrieval and template zone by the reproduced image in the 202 pairs of frames identical and different in matching part with the coded object piece.Then, to be set at prediction signal with the zone of the identical size of coded object piece by the portion of filling up 203, this coded object piece is with the position relation (in the example of Fig. 4, right side and downside) of regulation and the high correlation zone adjacency (S504, prediction signal deciding step) that obtains by template matches.The prediction signal that sets is imported into subtraction portion 102 and addition portion 106.
Then, generate difference signal (S505, coding step) by subtraction portion 102 by deducting the prediction signal of importing by prediction generating unit 108 (filling up portion 203) from original image by 101 inputs of Region Segmentation portion at pixel region.The difference signal that is generated is imported into transformation component 103, carries out conversion (S506) by transformation component 103.Difference signal after the conversion is imported into encoding section 104 and inverse transformation portion 105.The difference signal that is input to after the conversion of encoding section 104 carries out entropy coding by encoding section 104, generates compress coding data (S507, coding step).
The difference signal that is input to after the conversion of inverse transformation portion 105 carries out inverse transformation by inverse transformation portion 105, generates the difference signal (S508, reproduced image generates step) after the inverse transformation.Difference signal after the inverse transformation is imported into addition portion 106, is obtained and signal with the prediction signal addition of importing from prediction generating unit 108 (filling up portion 203) by addition portion 106, generates reproducing signal (S509, reproduced image generates step).The reproducing signal that is generated is imported into storage part 107, by storage part 107 be saved in frame memory etc. in the corresponding specified address (S510, storing step) of the frame that becomes coded object.
Next, judge (whether S502～S510) finishes (S511) in above-mentioned processing in all pieces of two field picture.Under situation about having finished, finish the processing of this frame, output is transferred to the processing of next frame by the compress coding data behind encoding section 104 entropy codings.When also leaving untreated, transfer to next piece and handle.In addition, above-mentioned judgement for example can be undertaken by above-mentioned any one inscape, and the unit that the encoding process in the blanket moving picture encoding device 100 also can be set carries out this unit.
In addition, in above-mentioned processing, be stored in reproducing signals in the storage part 107 and be based on by transformation component 103 conversion and by the difference signal after 105 inverse transformations of inverse transformation portion.This is because because the coding that carries out of supposition encoding section 104 is reversible, thereby can obtain the reproduced image that reproduces in above-mentioned processing in decoding device.
In above-mentioned processing, only enumerated by using the example of predicting of filling up of template matches, but also can carry out processing in addition.For example,, there is not the adjacent pixels of having reproduced sometimes,, is necessary to presuppose adjacent pixels in order to use this method in the upper end of two field picture or during the coding at two ends.In such part, code efficiency is better sometimes to use the prediction of motion vector as in the past.And, be not limited to the picture end, under the high situation of the code efficiency of comparing the prediction of using motion vector with the prediction of using this method, also can utilize the prediction of using motion vector.In this case, for example the information of which kind of mode by the mode that is to use expression prediction the manner also to be to use to utilize motion vector and motion vector value are newly given the head to block message, can separately make in two ways effectively.
As mentioned above,, when predicting between conducting frame,, do not use motion vector can carry out inter prediction, can carry out efficient coding by using the reproduced image of having encoded according to the moving picture encoding device 100 of present embodiment.That is, come down to the data that the difference signal after the only conversion is encoded, compare with existing coding method and deleted motion vector by the data behind moving picture encoding device 100 coding of present embodiment.And, in addition, when carrying out the decision of prediction signal, owing to only use the zone of having reproduced in this moment, thereby always can encode according to the scanning sequency of dynamic image data.
Below, the dynamic image decoding device of present embodiment is described.Fig. 6 illustrates the dynamic image decoding device 600 of present embodiment.Dynamic image decoding device 600 is to being decoded by the dynamic image data behind moving picture encoding device 100 codings and generating the device that reproduces dynamic image data.CPU), the hardware such as information processor of frame memory, hard disk etc. realize dynamic image decoding device 600 is by having CPU (Central Processing Unit:.The function inscape that dynamic image decoding device 600 is realized following explanation by above-mentioned hardware inscape is moved.
As shown in Figure 6, dynamic image decoding device 600 constitutes and has: lsb decoder 601, inverse transformation portion 602, addition portion 603, storage part 604 and prediction generating unit 605.Below, the function of each one is described.
Next, use the flow chart of Fig. 7 that the decoding processing (dynamic image decoding method) in the dynamic image decoding device 600 of present embodiment is described.Suppose that the compress coding data by dynamic image decoding device 600 decodings is by the data behind moving picture encoding device 100 codings.
At first, when the compress coding data of decoder object is imported into dynamic image decoding device 600, by lsb decoder 601 decode (S701, decoding step).Take out with the piece decoding back is the transform data of unit.This transform data is input to inverse transformation portion 602 by lsb decoder 601.And the positional information of decoder object piece in frame is imported into prediction generating unit 605 from the controller (not illustrating) of blanket decoding processing dynamic image decoding device 600.In addition, coded sequence is depended in the position of decoder object piece.
Then, the prediction signal that generates the decoder object piece by prediction generating unit 605 as described below.At first, by template zone determination section 201 with the reproduced image of piece adjacency on set template zone (S702, searching step).Then, carry out template matches, the high zone (S703, searching step) of reproducing signal correlation in retrieval and template zone by the reproduced image in the 202 pairs of frames identical and different in matching part with the coded object piece.Then, to be set at prediction signal with the zone of the identical size of coded object piece by the portion of filling up 203, this coded object piece is with the position relation (in the example of Fig. 4, right side and downside) of regulation and the high correlation zone adjacency (S704, prediction signal deciding step) that obtains by template matches.The prediction signal that sets is imported into addition portion 603.
Then, carry out inverse transformation, generate difference signal (S705, reproduced image generates step) by 602 pairs of transform datas of inverse transformation portion from lsb decoder 601 inputs.In addition, because a series of processing of S702～S704 and the processing of S705 can carry out before the later processing of the S706 of following explanation, thereby order can be opposite.
Then, will be obtained and signal with the prediction signal addition of importing from prediction generating unit 605 (filling up portion 203) from the difference signal of inverse transformation portion 602 inputs by addition portion 603, generating reproducing signal is decoding block (S706, reproduced image generates step).The decoding block that is generated is imported into storage part 604, by storage part 604 be saved in frame memory etc. in the corresponding specified address (S707, storing step) of the frame that becomes decoder object.
Next, judge (whether S701～S707) finishes (S708) in above-mentioned processing in all pieces of two field picture.Finishing and having finished under the situation of the two field picture that has reproduced, the two field picture that has reproduced is outputed to display units such as screen (not illustrating).And, also can output to other demonstration device.When also leaving untreated, transfer to the processing of next piece.In addition, above-mentioned judgement for example can be undertaken by above-mentioned any one inscape, and the unit that the encoding process in the blanket dynamic image decoding device 600 also can be set carries out this unit.
As mentioned above, according to the dynamic image decoding device 600 of present embodiment, can determine prediction signal to come dynamic image is decoded in the same manner with moving picture encoding device 100.That is,, can be correctly decoded the dynamic image data that has carried out efficient coding by above-mentioned moving picture encoding device 100 and generate reproduced image according to the dynamic image decoding device 600 of present embodiment.
In addition, can consider following variation in the present embodiment.Following variation is described moving picture encoding device and dynamic image decoding device, but can implement too for moving picture encoding processing and moving image decoding processing.
(1) range of search
So far, be that range of search is made as the reproduction regions (403) and the reproduced image 404 of rendering frame in the coded object frame with the template matches object, but also can be that any one party is selected by unit with the frame.For example, if be limited to reproduction regions 403, then the present invention can also be applied to only the reproducing signal in the frame is made as in the frame of forecasting object.By only reproduced image 404 being defined as searching object, has the effect of the operand in the time of reducing the interframe decoding of following the inter prediction generation equally.And, prepare and be that unit selects that reproduction regions 403 is made as the predictive mode (coupling prediction in the template) of the searching object of template matches and reproduced image 404 is made as the also reduction of operand owing to can realize decoding the time of method of the predictive mode (coupling is predicted between template) of the searching object of template matches with the piece, thereby be effective.At this moment, shown in the explanation of Fig. 5, use can also combine with a plurality of predictive modes of non-patent literature 1 record.In addition, the range of search in each image is made as predefined zonule also owing to can realize the minimizing of operand, thereby is effective.
(2) multiframe
So far, although the number as the rendering frame of template matches object is not done special description, yet be made as object owing to can improve estimated performance with many, thereby be effective.At this moment, can comprise the interior reproduced image zone of coded object frame.Below, with these reproduced image zones that become the reproduced image of rendering frame of template matches object and coded object frame be referred to as reference frame with reference to image.At this moment, can use the high-resolution image that not only comprises integer pixel but also comprise the real number pixel that generates by Filtering Processing with reference to image.The generation method of real number pixel for example is documented in the non-patent literature 1.The reference frame selected numbering can be unit with the piece or be that unit encodes with the frame, also can have additional information ground according to desired values such as SAD and not select.Under the situation of encoding expressly, original signal that will (in the coded object piece) forecasting object zone is compared with the prediction signal according to (in the coded object piece) forecasting object zones that generates with reference to image of a plurality of reference frame, selects 1 reference frame.
This encoding process can be implemented by append selection portion 109 between prediction generating unit 108 and subtraction portion 102 in moving picture encoding device shown in Figure 1 100 as shown in figure 18.Selection portion 109 is at being the prediction signal that object generates by prediction generating unit 108 with a plurality of reference frame, calculates the desired value (SAD or MSE etc.) between the original signal with the coded object piece, the reference frame of selection desired value minimum.The reference frame numbering of selecting is carried out entropy coding by encoding section 104.In addition, the pack processing of selection portion 109 is contained in the prediction generating unit 108 also can obtain identical result, thereby this structure also can be implemented this variation.In decoding processing, as shown in figure 19, prediction generating unit 606 is utilized and is generated prediction signal by the corresponding reference frame of lsb decoder 601 decoded reference frame numberings.
And, can not to select 1 reference frame to generate the prediction signal (1 frame is selected to handle) of coded object piece, but select 2 reference frame and to be unit with the pixel calculate final prediction signal (equalization processing) to the averaging of prediction signal of the coded object piece obtained from the reference frame of selecting.At this moment, also can consider to obtain the selection candidate of the prediction signal of coded object piece from identical reference frame.When selecting the prediction signal of 2 positions that have been offset 1 pixel (or 1/2 pixel or 1/4 pixel) from identical reference frame, also has the effect of handling the precision that improves the motion vector that to retrieve by interpolation.In addition, smoothing is handled owing to the effect with noise component of removing predictive error signal, thereby the conformability of general and transition coding is good.
Under the situation of selecting 2 reference frame, can not only can also calculate the final prediction signal (weighted average processing) in forecasting object zone by the weighted average processing of each pixel by simple equalization.Weight coefficient has the method for using steady state value in piece or the method that changes as unit with pixel.Establishing method or coding method about weight coefficient are not particularly limited, and can use the method for non-patent literature 1 record.
And, select reference frame that odd number opens in addition and be the method (median prediction processing) of the median of the unit prediction signal of selecting the forecasting object zone with the pixel.And, can also implement such method, that is: be the prediction signal central authorities in forecasting object zone are selected by unit from the frame that the even number more than 4 is opened 2 values with the pixel, and calculate predicted value by equalization or weighted averageization.The Forecasting Methodology of utilizing median like this has the effect that reduces the consensus forecast error.
These equalizations processing, weighted average processing and median prediction processing can be realized (being applicable to the prediction generating unit 108 of Fig. 1 and the prediction generating unit 605 of Fig. 6) by the prediction generating unit 1108 after appending signal generating unit 204 to prediction generating unit 108 shown in Figure 2 as shown in figure 20.In signal generating unit 204, the prediction signal in the forecasting object zone that is generated by multiframe is made as input, use the processing method shown in above-mentioned to generate final prediction signal.Because the processing of signal generating unit 204 can use the information that belongs to the characteristic (motion vector etc.) of reproduced image and derive according to reproduced image or (as the data relevant with reproduced image) to implement, thereby also can implement same treatment in the decoding side.
Also can consider such method, that is: select a plurality of processing methods of preparation such as processings, equalization processings, weighted average processing, median prediction processing according to 1 frame, and be unit with the piece or be unit selection processing method with the frame.In the prediction processing of using the template that constitutes by decode value, do not guarantee that the motion of desired value the best minimizes predictive error signal.Therefore, it is effective selecting the method for proper method from the different a plurality of processing methods of the prediction signal characteristic in forecasting object zone.
As the system of selection of processing method, can consider such method, that is: at first select to make the predicated error absolute value and (perhaps squared prediction error and) the minimum method (selection portion 109 of Figure 18) in forecasting object zone, and be sent to the decoding side in the coding side.The prediction generating unit 1108 that this system of selection can be replaced as Figure 20 by the prediction generating unit 606 with the prediction generating unit 108 of Figure 18 and Figure 19 realizes.In addition, the selection portion 109 of Figure 18 outputs to encoding section 104 to the information of selected processing method here, and replaces selected reference frame numbering.Because the structure that the pack processing of selection portion 109 is contained in the prediction generating unit 1108 also can obtain identical result, thereby this structure also can be implemented.
And, also can consider such method, that is: generate the prediction signal in template zone according to separately processing method, and and the generation signal in template zone between parameter value (SAD or MSE etc.) select processing method automatically.For example, in equalization is handled, be that unit calculates prediction signal to the averaging of candidate of 2 prediction signal in template zone with the pixel, parameter value between the reproducing signal in prediction signal that calculates and template zone.According to this method, can use according to reproduced image or the information that belongs to the characteristic (motion vector etc.) of reproduced image and derive determines processing method uniquely, thereby need not the information of processing method is encoded.
And, also can consider such method, that is: variance yields (TaV) by the signal in the template zone and combination at the variance yields (EvV) of the desired value of a plurality of reference frame, set 4 conditions, use equalization to handle at TaV and EvV under all less than the situation of threshold value, use the weighted average processing under less than the situation of threshold value, use the median prediction processing under less than the situation of threshold value at EvV only at TaV only, all under the situation greater than threshold value, use 1 frame to select to handle at TaV and EvV.In this case, estimate the characteristic strength of the direction in space of template according to the variance of the reproducing signal in template zone, and estimate the characteristic strength of the time orientation in template zone according to the variance of desired value.In the method, implement the selection of processing method according to such supposition, that is: variance yields is more little, and the difference of a plurality of candidates in the prediction signal in forecasting object zone is just more little, and the effect of noise of removing the predictive error signal that smoothing causes is just high more.This automatic selecting method can be replaced as by prediction generating unit 108 Figure 18 Figure 20 prediction generating unit 1108, and the system of selection shown in above-mentioned is imported in the selection portion 109 of Figure 18 and realizes.In this system of selection, also can use according to reproduced image or the information that belongs to the characteristic (variance yields etc.) of reproduced image and derive determines processing method uniquely, thereby need not the information of processing method is encoded.Therefore, omitted from the output of selection portion 109 to encoding section 104.
At this moment, in the system of selection of the user's difference shown in above-mentioned, what need is: being replaced as the reproducing signal in template zone to the input of selection portion 109 from the original signal of coded object piece, and the desired value of a plurality of reference frame is appended to from predicting the input of generating unit 1108 to selection portion 109.In addition, also can obtain identical result because prediction generating unit 1108 comprises the structure of the processing of selection portion 109, thereby this structure can be implemented also.The processing of decoding side can be by according to encoding process, and the combination that the prediction generating unit 605 of Fig. 6 is replaced as the selection portion 109 of the prediction generating unit 1108 of Figure 20 and Figure 18 realizes.In addition, the method that automatic selecting method is not limited to put down in writing here is so long as only utilize according to reproduced image or the method that belongs to the information that the characteristic of reproduced image derives just can realize.
(3) structure of prediction generating unit
(4) size of coded object piece
The size of coded object piece is made as 8 pixels * 8 pixels in the present embodiment, yet even other piece size of the present invention also can be implemented, thereby is not limited to this size.And, the size in template zone is not limited yet.For example, following various situation is arranged, that is: template zone and forecasting object zone are added obtain 12 pixels * 12 pixels together, the size in template zone is got 10 pixels * 10 pixels of half etc. for 8 pixels * 8 pixels.And, be unit with the piece or be that the size of unit change coded object piece and the size in template zone also are effective with the frame.Shown in non-patent literature 1, prepare the coded object piece of different size and the group in template zone and can tackle the interior different pattern of image, thereby be effective.And, in considering template between coupling prediction and template under the situation of coupling prediction, when piece size that coupling in the general low template of the redundancy that reduces template zone and range of search is predicted, can expect the raising of forecasting efficiency.
And,, also can implement prediction of the present invention even the forecasting object piece is different piece sizes with the coded object piece.
(5) reproducing signal in template zone
So far, the reproducing signal in template zone is made of the reproduction pixel of coded object frame, yet the pixel of reproducing on other reference frame is known in the decoding side also, thereby can utilize.For example, consider such situation, that is: the forecasting object Region Segmentation with 8 pixels * 8 block of pixels becomes 4 pixels * 4 block of pixels, is unit generation forecast signal with 4 * 4.When template zone and forecasting object zone being added obtain 6 pixels * 6 block of pixels, 4 * 4 upper left template zone for 8 * 8 can be made of the reproduction pixel of coded object frame together.Yet in 8 * 84 * 4 upper right template zone, 6 * 2 pixels on piece top can be made of the reproduction pixel of coded object frame, yet owing to 2 * 2 pixels of piece left part are also encoded, thereby replace with 4 * 4 upper left prediction signal.Equally, for all pixels in 4 * 4 template zones of 2 * 2 pixels on the piece top among 4 * 4 of 8 * 8 lower-lefts and bottom right, because the reproduction pixel of coded object frame coding as yet, thereby replace with prediction signal.
And, so far obtain with the piece of coded object piece adjacency reproduce the reproducing signal that pixel constitutes the template zone, but also can generate the reproducing signal in template zone to the filtering etc. that reproducing signal implements to remove noise.For example, in the many images of noise, can be by implementing the motion detection that Filtering Processing does not have noise effect to the reproducing signal in template zone with reference to image.
(6) about desired value
So far, desired value during by the prediction signal in the template matches formation object zone is used the absolute difference of prediction signal in the template zone and object signal and (SAD) or Mean Square Error (MSE), yet is not limited thereto.For example, as T.Wiegand et.al. " Rate-ConstrainedCoder Control and Comparison of Video Coding Standards ", IEEE Trans.Circuits and Systems for Video Technology, vol.13, No.7, July 2003,688-703. (non-patent literature 2) put down in writing like that, the value after the size of difference motion vector taken in also can be used as desired value of the present invention.
And it also is effective that the absolute difference of each pixel or square error are weighted.For example, can consider such method, that is, SAD is being made as under the situation of evaluation of estimate, to the absolute difference setting 4 of the boundary pixel of template zone and estimation range, be 3,2,1 whenever leaving the border with regard to reducing weight.Preferential by making like this near the pixel on border, can expect the raising of estimated performance.
(7) structure of inverse transformation portion
In Fig. 1, the input of inverse transformation portion 105 becomes the output from transformation component 103, but also can be the output from encoding section 104.In this case, before the processing of inverse transformation portion 105, implement the processing of the lsb decoder 601 of Fig. 6.In addition, the processing of moving picture encoding device and dynamic image decoding device is being carried out also implementing the present invention in the integrated installation method.That is, also can consider such structure, that is: handle by the output of 601 pairs of encoding section 104 of lsb decoder of Fig. 6, and decoded picture is input to storage part 107.
[the 2nd execution mode]
The moving picture encoding device in the 2nd execution mode and the apparatus structure of dynamic image decoding device are except the detailed structure of prediction generating unit, and be identical with apparatus structure in the 1st execution mode.Below, the difference of prediction generating unit in the present embodiment and the prediction generating unit in the 1st execution mode 108,605 is described.
As shown in Figure 8, the prediction generating unit 800 of present embodiment constitutes and has: judging part 801, template zone determination section 802, matching part 803 and fill up portion 804.
Judging part 801 be with the reproducing signal in template zone with compare by the reproduced image of storage part 107,604 storages, and estimate to become the estimation unit of spatial continuity of signal in the zone (forecasting object piece) of coding or decoder object according to this comparative result.In addition, spatial continuity is the indexs of features such as the travel direction of expression in certain space, size in which kind of degree unanimity.That is, in certain zone, under this regional the first half situation different, there is not spatial continuity with the latter half motion feature.And, judging part 801 is the spatial continuities according to the image that estimates, further be partitioned into the zone of coding or decoder object, this divided area is set at the zone (forecasting object zone, fill up zone) of new coding or decoder object, and sets setup unit at the template zone in the zone of this new coding or decoder object.801 pairs of judging parts are analyzed by the reproduced images of storage part 107,604 storages, and decision comprises the Prediction Parameters of the size in the size in template zone and forecasting object zone, and its information is outputed to template zone determination section 802 and fills up portion 804.The concrete decision method of Prediction Parameters is described in the back.
Template zone determination section 802 is according to the dimension information from the template zone of judging part 801 inputs, sets an inscape of the retrieval unit of the template zone that is used for the generation forecast signal and this regional image.Template zone determination section 802 is corresponding to the zone determination section 201 of the template in the 1st execution mode and have identical function.
Matching part 803 is to use the image in the template zone that is set by template zone determination section 802, implementing template matches as search domain by the reproduced image of storage part 107,604 storage, and the retrieval unit in the zone that retrieval and the correlation of the pixel groups in template zone are the highest in search domain.Matching part 803 is corresponding to the matching part in the 1st execution mode 202 and have identical function.
Fill up portion 804 and be based on the zone (high correlation zone) that retrieves by matching part 803 and the relation of the position between forecasting object piece and the template zone, determine the unit according to the prediction signal of the prediction signal of reproduced image setting and decision and forecasting object piece same size.The size of the forecasting object piece of this moment is the size that is set by judging part 801.Fill up portion 804 corresponding to the portion of filling up 203 in the 1st execution mode and have identical function.
Here, use Figure 14 and Figure 15 that the determining method of the size in the size in the template zone in the judging part 801 and forecasting object zone is described.When the size in size that determines the template zone and forecasting object zone, utilize reproduction pixel with forecasting object piece adjacency.Figure 14 illustrates the pixel of forecasting object piece 1401 and the figure that reproduces pixel 1402 of its periphery.At first, as Figure 14 (a) with (b), prepare following 4 zones, promptly; Cover the regional A with the zone 1402 of forecasting object piece 1401 adjacency fully, and zone 1402 is cut apart and the area B of the part of overlay area 1402, C, D (area B, C, D do not overlap separately, add form be zone 1402) respectively together.
Then, implement regional A, B, C, D as the template zone at template matches by the reproduced image of storage portion stores, and obtain the high zone of correlation respectively.Here, as relevance values, suppose and use for example SAD.SAD at each regional A, B, C, D is made as SAD
A, SAD
B, SAD
C, SAD
D, with SAD
AWith (SAD
B+ SAD
C+ SAD
D) compare.At SAD
AUnder the very large situation, being estimated as in regional A does not have spatial continuity, and does not have spatial continuity in forecasting object piece 1401 yet.That is, as shown in figure 15, being estimated as is to lay respectively at state in the zone 1501 of reproduced image with the high regional 1501a of regional A correlation and with area B, C, regional 1501b, 1501c, 1501d that the D correlation is high.Estimate according to this, be judged as in the template matches of usage space A prediction and carry out smoothly (departing between prediction signal that is determined and the Code And Decode object images increase).SAD
AVery large judgement can wait and carries out by being predetermined threshold value.After judging like this, further block is set the size in the size in template zone and forecasting object zone forr a short time than the size of piece.The size in the forecasting object zone of this moment for example can use and the corresponding size of cutting apart of area B, C, D.And the size corresponding with the size in this forecasting object zone used in the template zone.
Otherwise at SAD
AWith (SAD
B+ SAD
C+ SAD
D) under the little discrepant situation, be estimated as the continuity of in regional A, having living space, and the continuity of in forecasting object piece 1401, also having living space.Estimate that according to this being judged as the template matches of using regional A is effectively, the size (with the forecasting object zone as the not divided zone of piece) of piece is used in the forecasting object zone.In addition, be not only the such figure of area B, C, D shown in Figure 14 (b), can also shown in Figure 14 (c), further segment carrying out divided area with the zone 1402 of forecasting object piece 1401 adjacency of in above-mentioned judgement, using.
In the prediction of using the such template matches of the present invention, different with the prediction of using motion vector, can not reach prediction accurately.Therefore, be necessary to avoid as far as possible misprediction.And, it is generally acknowledged the size hour easy misprediction that takes place when the template zone.On the other hand, trickle and do not have in the part of spatial continuity in motion, when template zone or forecasting object zone are big, can not tackle slight movement, predicated error increases.Therefore, as use the method for the present invention, the size that reduces the size in template zone and forecasting object zone improves the probability that is suitable for slight movement and becomes effective.
Prediction when using Fig. 9 to the size in the size that changed the template zone and forecasting object zone is passed and is described.Having the size in spatial continuity and forecasting object zone in the zone with forecasting object piece 901 adjacency is under the situation of forecasting object piece 901 integral body, handles by the template matches of 1 degree in the same manner with the 1st execution mode.For example, as shown in Figure 9,, reduce the size in forecasting object zone, and the also corresponding situation that reduces the size in template zone describes to forecasting object piece 901 being divided into 4 zones of regional 901a, 901b, 901c, 901d.In addition, suppose that the left side of forecasting object piece 901 and the zone of upside are the image-regions that has reproduced.
At first, shown in Fig. 9 (a),, the regional 902a of left side and upside is set at the template zone, uses the prediction signal of template matches to set at the upper left regional 901a of forecasting object piece 901.Then, shown in Fig. 9 (b),, the regional 902b of upside is set at the template zone, uses the prediction signal of template matches to set at the regional 901b on regional 901a right side.Then, shown in Fig. 9 (c), the regional 901c at the regional 901a downside that has carried out the prediction signal setting at first is set at the template zone to the regional 902c in left side, uses the prediction signal of template matches to set.Then, shown in Fig. 9 (d), at remaining area is the regional 901d of bottom right, and the regional 902d of the left side of inclusion region 901a, 901b, 901c and upside is set at the template zone, uses prediction signal the prediction signal of template matches to set as the object signal in template zone.Thus, set prediction signal, can carry out Code And Decode at the All Ranges of forecasting object piece 901.
In addition,, not only cut apart, can also shown in Fig. 9 (e), (f), only cut apart in any one direction in length and breadth as the both direction of Fig. 9 (a)～(d) be shown in length and breadth as the variation of the size in the size in template zone and forecasting object zone.For example, in Figure 14 (b), include in the high correlation zone in regional A among area B, C, D, the E the high correlation zone rather than only under the situation in the high correlation zone among the inclusion region F, can be shown in vertically and cut apart as Fig. 9 (e), (f).This is because the first half and the latter half that can be judged as the forecasting object piece in this case do not have spatial continuity.
In this case, at first, shown in Fig. 9 (e),, the regional 902e of left side and upside is set at the template zone, uses the prediction signal of template matches to set at the regional 901e of forecasting object piece 901 the first half.Then, shown in Fig. 9 (f),, the regional 902f in left side is set at the template zone, uses the prediction signal of template matches to set at the regional 901f of regional 901e downside.Thus, set prediction signal, can carry out Code And Decode at the All Ranges of forecasting object piece 901.
Next, use the flow chart of Figure 10 that the encoding process in the moving picture encoding device of present embodiment (motion image encoding method) is described.
At first, when the dynamic image data of coded object is imported into moving picture encoding device, be input to Region Segmentation portion 101 according to each two field picture that constitutes dynamic image.The two field picture of being imported is divided into a plurality of (S1001) of prescribed level by Region Segmentation portion 101.Later processing all is that unit carries out with the piece.Piece is imported into prediction generating unit 800 and subtraction portion 102 as the pixel in coded object zone.
Then, the prediction signal that generates the coded object piece by prediction generating unit 800 as described below.The pixel of reproducing by judging part 801 uses and coded object piece adjacency decides the Prediction Parameters of using (S1002, estimating step and determining step) in this coded object piece at first, as mentioned above.The Prediction Parameters that is determined is imported into template zone determination section 802.Then, set the coded object zone according to the Prediction Parameters that sets, will be set at template (S1003) reproducing on the pixel region (template zone) with the reproduction pixel groups of this zone adjacency by template zone determination section 802.In addition, the coded object zone that sets here is as mentioned above the coded object piece have been carried out divided area sometimes.Then, carry out template matches, the high zone (S1004) of pixel groups correlation in retrieval and template zone by the reproduced image in the 803 pairs of frames identical and different in matching part with the coded object piece.Then, by the portion of filling up 804 will with the position relation of regulation and the high correlation zone that obtains by template matches in abutting connection with and be set at prediction signal (S1005) with the zone of the regional identical size of coded object.
Then, whether the All Ranges of judging the coded object piece has all set prediction signal (S1006).All set under the situation of prediction signal in not all zone, repeated a series of processing of S1003～S1005, till the All Ranges of coded object piece has all been set prediction signal.In addition, above-mentioned judgement for example can be undertaken by above-mentioned any one inscape, and the unit that the encoding process in the blanket moving picture encoding device also can be set carries out this unit.
Following processing (corresponding processing (the S505～S511) identical in S1007～S1013) and the 1st execution mode.
As mentioned above, identical according to the moving picture encoding device of present embodiment with the 1st execution mode, when predicting between conducting frame,, do not use motion vector can carry out inter prediction by using the reproduced image of having encoded, can carry out efficient coding.And, owing to can suitably select the size in the zone of prediction signal according to the spatial continuity of reproducing signal, even thereby change in the coding of violent dynamic image data at amount of exercise, also can improve code efficiency.
Next, use the flow chart of Figure 11 that the decoding processing in the dynamic image decoding device of present embodiment (dynamic image decoding method) is described.Suppose that the compress coding data of being decoded by the dynamic image decoding device of present embodiment is by the data behind the moving picture encoding device coding of above-mentioned present embodiment.
At first, when the compress coding data of decoder object is imported into dynamic image decoding device, by lsb decoder 601 decode (S1101).Take out with the piece decoding back is the transform data of unit.This transform data is input to inverse transformation portion 602 by lsb decoder 601.And the positional information of decoder object piece in frame is input to prediction generating unit 800 by lsb decoder 601.In addition, coded sequence is depended in the position of decoder object piece.
Then, the prediction signal that generates the decoder object piece by prediction generating unit 800 as described below.Reproduced image by judging part 801 uses and decoder object piece adjacency decides the Prediction Parameters of using (S1102, estimating step and determining step) in this decoder object piece at first, as mentioned above.The Prediction Parameters that is determined is imported into template zone determination section 802.Then, according to the Prediction Parameters setting decoder object zone that sets, will be set at template (S1103) by template zone determination section 802 with the reproduction pixel groups of this zone adjacency.In addition, the decoder object zone that sets here is the zone that as mentioned above the decoder object piece is split to form sometimes.Then, carry out template matches, the high zone (S1104) of pixel groups correlation in retrieval and template zone by the reproduced image in the 803 pairs of frames identical and different in matching part with the decoder object piece.Then, by the portion of filling up 804 will with the position relation of regulation and the high correlation zone that obtains by template matches in abutting connection with and be set at prediction signal (S1105) with the zone of the regional identical size of decoder object.
Then, whether the All Ranges of judging the decoder object piece has all set prediction signal (S1106).All set under the situation of prediction signal in not all zone, repeated a series of processing of S1103～S1105, till the All Ranges of decoder object piece has all been set prediction signal.In addition, above-mentioned judgement for example can be undertaken by above-mentioned any one inscape, and the unit that the decoding processing in the blanket dynamic image decoding device also can be set carries out this unit.
Following processing (corresponding processing (the S705～S708) identical in S1107～S1110) and the 1st execution mode.In addition, identical with the 1st execution mode, a series of processing of S1102～S1106 and the processing of S1107 can be carried out before the later processing of S1108, thereby order can be opposite.
As mentioned above, according to the dynamic image decoding device of present embodiment, can determine prediction signal to come dynamic image is decoded in the same manner with the moving picture encoding device of present embodiment.That is,, can carry out the dynamic image data after the efficient coding to above-mentioned moving picture encoding device and be correctly decoded and generate reproduced image by present embodiment according to the dynamic image decoding device of present embodiment.
In addition, can consider following variation in the present embodiment.Following variation is described moving picture encoding device and dynamic image decoding device, yet can implement too for moving picture encoding processing and moving image decoding processing.
(1) template zone and forecasting object zone separates
As mentioned above, determine the size and dimension in forecasting object zone and template zone simultaneously by the judging part 801 of Fig. 8, yet a side size and dimension is fixed.Therefore, fix and want to switch adaptively under the situation of the size in template zone or shape, also can use the present invention at the size and dimension that makes the forecasting object zone.In this case, do not need from the output of judging part 801 to the portion of filling up 804.For example, in Figure 14, can consider to select the regional A, area B, zone C, region D from Figure 14 (a) and (b) method at the template in forecasting object zone 1401.As the template in forecasting object zone 1401, and have the pattern continuity between the forecasting object zone 1401 and constitute the pixel count of this template can be many.Therefore, common regional A is preferred, yet does not have can use undersized template under the successional situation of pattern in certain direction.In addition, the determination processing in the judging part 801 is not limited to the method shown in above-mentioned.For example, can consider such method, that is: the prediction signal in template zone be compared with actual prediction signal after this regional pixel is encoded, select little shape of template and the size of mean difference absolute value.Can also use the prediction signal after the reproducing signal of the coded object piece under this zone rather than use are encoded to regenerate prediction signal once more.And, also can consider such method, that is: select the prediction signal in template zone and the little shape of template and the size of mean difference absolute value of object signal (reproducing signal).
On the other hand, guarantee motion continuity rather than guarantee that method of patterning also is effective.For example, calculate regional A as the template zone and the motion vector of detected motion vector and adjacent block or poor according between the motion vectors of adjacent block calculating.Then, under the situation of motion vector less than predetermined threshold value of difference, the motion vector of detected motion vector as the estimation range.On the other hand, under the situation of motion vector greater than predetermined threshold value of difference, detect motion vector at different shape of template (for example, area B, D, the F etc. of area B, C, D or Figure 14 (c) of Figure 14 (b)).And, also can consider such method, that is: at each template zone, the motion vector in the time of will encoding to the pixel in this zone is compared with detected motion vector, the little motion vector of difference is chosen as the motion vector in forecasting object zone.Can also substitute the motion vector the when pixel in this zone encoded, and use the reproducing signal of the coded object piece under this zone to come to detect again once more motion vector.No matter in which method, can both use the information that belongs to the characteristic (motion vector etc.) of reproducing signal and derive from reproducing signal or (as the information relevant) to implement with reproducing signal.And, can also the size of detected motion vector be compared according to a plurality of shape of template and decide the size of shape of template and estimation range.The size of the motion vector of adjacent block can also be compared equally and decide the size of shape of template and estimation range.
The shape and size that become candidate's template are not limited to template shown in Figure 14 and cut apart.For example, following situation is also contained in this variation, that is: for 8 pixels * 8 pixels, select to make template zone and forecasting object zone to add together size from 12 pixels * 12 pixels, 10 pixels * 10 pixels and 14 pixels * 14 pixels.
(2) the size decision in template zone and forecasting object zone
As mentioned above, by the judging part 801 of Fig. 8 by calculating the size and dimension in decision template zone and forecasting object zone, yet also can utilize the information of the next size and dimension to the best of the original signal (coded object piece) in forecasting object zone to encode.This determining method for example can be implemented in the following manner, that is: output predicts as shown in figure 18 that generating unit 108 is used multiple templates and the prediction signal in the forecasting object zone that generates, make desired value (SAD, MSE etc.) minimum template size and shape by the selection portion selection, carry out entropy coding by 104 pairs of these information of encoding section.The structure that the pack processing of selection portion 109 is contained in the prediction generating unit 108 also can be implemented.
And,, do not have the object signal in template zone sometimes according to the position in forecasting object zone.In this case, only generate template according to the object signal that exists.For example, the zone C shown in Figure 14 (b) is made as the template zone, region D is made as the template zone in the upper end of image at the left end of image.
[the 3rd execution mode]
In the motion prediction of the present invention that uses the template constitute by reproducing signal, detect the motion of the desired value minimum of the difference signal that object signal and prediction signal according to the template zone calculate from range of search.Therefore, the object signal in the template zone has under the situation of notable feature, can implement suitable motion prediction according to this feature.Yet as flat site, the object signal in template zone does not show under the situation of notable feature, even the desired value minimum detects with the possibility of actual different motions and also increases.In this case, when the prediction signal in forecasting object zone and the difference between the object signal were big, its sign indicating number amount also increased.On the other hand, do not have the flat site of notable feature, the high fdrequency component that comprises in the signal is few, has reduced the signal of resolution and the space similitude height between the primary signal.Therefore, reduce the resolution of the object signal in forecasting object zone and encode, and use straightforward procedure that the reproducing signal that has reduced resolution is expanded, also can suppress degeneration from primary signal in the decoding side.Therefore, in the present embodiment, as the difference coding method that is suitable for flat site, the resolution of the object signal that reduces the forecasting object zone and prediction signal is shown and this low resolution difference signal is carried out Methods for Coding (dwindle coded object piece and prediction piece and it is dwindled difference block carry out Methods for Coding).
In addition, in the present embodiment, for convenience's sake, the piece that is made of prediction signal is called the prediction piece, the piece that is made of difference signal is called difference block, the piece that is made of reproducing signal is called decoding block.And the coded object piece is that the piece that the original signal by the coded object frame of dynamic image data constitutes is called the coded object piece.
Figure 21 illustrates the moving picture encoding device 2100 of implementing the 3rd execution mode.Can realize by portion 2110 that dwindles (general name of dwindling the 2110-1 of portion and dwindling the 2110-2 of portion) and enlarging section 2111 are set in the moving picture encoding device 100 of Fig. 1.In addition, the function of subtraction portion 2102, transformation component 2103, encoding section 2104, inverse transformation portion 2105 and addition portion 2106 just reduces to be processed size (for example 4 pixels * 4 pixels), and the function with subtraction portion 102, transformation component 103, encoding section 104, inverse transformation portion 105 and the addition portion 106 of Fig. 1 is identical respectively.In addition, about transformation component 2103 and inverse transformation portion 2015, in the transformation component 103 of Fig. 1 and inverse transformation portion 105, also shown in non-patent literature 1, can be that unit handles with 4 pixels * 4 pixels.In this case, the transformation component 103 of Fig. 1 is meant that with transformation component 2103 and the different of inverse transformation portion 2105 of Figure 21 to be processed quantity reduces to 1 from 4 with inverse transformation portion 105.
Dwindle the 2110-1 of portion and dwindle coded object piece that the 2110-2 of portion will obtain from Region Segmentation portion 108 respectively and dwindle into and dwindle the coded object piece and dwindle the prediction piece, and output to subtraction portion 2102 from the prediction piece that prediction generating unit 108 obtains.In subtraction portion 2102, be that unit calculates 2 differences of dwindling piece with the pixel, and output to transformation component 2103 dwindling difference block.In transformation component 2103, implement conversion (and quantization) and handle, transform data (quantization data) portion's 2104 codings that are encoded.Simultaneously, transform data (quantization data) is implemented (inverse guantization (IQ) and) inversion process by inverse transformation portion, and difference block is dwindled in decoding outputs to addition portion 2106.In addition portion 2106, be that unit dwindles difference block with decoding and dwindles the addition of prediction piece with the pixel, generate decoding and dwindle piece.Enlarging section 2111 dwindles piece with decoding and is amplified to decoding block with coded object piece same size, and outputs to recording portion 107.Dwindling the processing of portion 2110 and enlarging section 2111 uses Figure 23 to describe in the back.
Figure 22 illustrates the dynamic image decoding device 2200 of implementing the 3rd execution mode.Can dwindle portion 2207 and enlarging section 2208 is realized by in the dynamic image decoding device 600 of Fig. 6, being provided with.The processing of dwindling portion 2207 and enlarging section 2208 have respectively with Figure 21 in portion that dwindles 2110 and enlarging section 2111 identical functions.In addition, the function of lsb decoder 2201, inverse transformation portion 2202 and addition portion 2203 just reduces to be processed size (for example 4 pixels * 4 pixels), and the function with lsb decoder 601, inverse transformation portion 602 and the addition portion 603 of Fig. 6 is identical respectively.In addition, inverse transformation portion is identical with the code device of Fig. 1, also shown in non-patent literature 1, can be that unit handles with 4 pixels * 4 pixels in the inverse transformation portion 602 of Fig. 6.In this case, the inverse transformation portion 602 of Fig. 6 is meant that with the different of inverse transformation portion 2202 of Figure 22 to be processed quantity reduces to 1 from 4.
Dwindle portion 2207 and will dwindle into from the prediction piece that obtains of prediction generating unit 605 and dwindle the prediction piece, and output to addition portion 2203.In addition portion 2203, be that unit will decoded decoding dwindle piece and dwindle the addition of prediction piece by the processing of lsb decoder 2201 and inverse transformation portion 2202 with the pixel, generate decoding and dwindle piece.Identical with the code device of Figure 22, inverse transformation portion 2202 can comprise quantization and handle.Enlarging section 2208 dwindles piece with decoding and is amplified to decoding block with the decoding block same size, and outputs to recording portion 604.
Figure 23 illustrates dwindling/processing and amplifying in portion of dwindling 2110,2207 and the enlarging section 2111,2208.Piece before piece 2301 expressions are reduced.Handling 2304 pairs of pixel generation methods of dwindling on the piece of dwindling in the processing describes.Pixel j, k, m, n represent the pixel on the piece 2301, and pixel P represents to be generated dwindles pixel on the piece.Handling in 2304, be the equalization processing that unit implements pixel with 4 pixels, and calculate the pixel of dwindling on the piece.Piece 2302 expressions are handled the piece that dwindles that obtains by dwindling.
Handle the pixel generation method on the amplification piece in the 2305 expression processing and amplifying.Pixel A～D represents the pixel on the piece 2302, and pixel a～i represents the pixel on the enlarged image.Handling in 2305, using diverse ways to implement the interpolation/extrapolation process of pixel according to location of pixels.Pixel a is because the pixel of dwindling on the piece of adjacency only is A, thus with pixel A directly as pixel a.Equally, the pixel of being represented by white circle of piece 2303 is used and is duplicated the method for dwindling the adjacent pixels on the piece and calculate.About pixel b～e, there are 2 pixels in the pixel of dwindling on the piece of adjacency.Therefore, these pixels extrapolation process of dwindling adjacency 2 pixels on the piece by use is calculated.Equally, the pixel of being represented by black box of piece 2303 extrapolation process of dwindling adjacency 2 pixels on the piece by use is calculated.About pixel f～i, there are 4 pixels in the pixel of dwindling on the piece of adjacency.Therefore, the linear interpolation of dwindling adjacency 4 pixels on the piece by use of these pixels is handled and is calculated.Equally, the linear interpolation of dwindling adjacency 4 pixels on the piece by use of the pixel of being represented by black circle of piece 2303 is handled and is calculated.
Figure 24 and Figure 25 illustrate moving picture encoding processing and the moving image decoding of realizing the 3rd execution mode respectively and handle.Figure 24 and Figure 25 correspond respectively to Fig. 5 and the Fig. 7 in the 1st execution mode.In Figure 24 and Figure 25, (S502 is handled in the decision of template among Fig. 5 and Fig. 7 zone, S702), template matches handle (S503, S703), the prediction signal decision handles (S504) and prediction signal and generate and handle (S704) and accumulate prediction signal and generate and handle that (S2404 S2504) puts down in writing.And the S2401 among Figure 24, S2405, S2406, S2407, S2408, S2410, S2411 correspond respectively to S501, S505, S506, S507, S508, S510, the S511 among Fig. 5.These are handled and just make to be processed size difference, owing to implement identical processing, thereby omit explanation here.And the S2501 among Figure 25, S2505, S2507, S2508 correspond respectively to S701, S705, S707, the S708 among Fig. 7.These are handled and also just make to be processed size difference, owing to implement identical processing, thereby omit explanation here.
Encoding block in the treatment S 2412 of Figure 24 dwindles in the processing, will be dwindled into from the coded object piece of Region Segmentation portion 101 inputs by the portion 2110-1 of dwindling and dwindle the coded object piece, and the piece after dwindling is outputed to subtraction portion 2102.Prediction piece in treatment S 2313 dwindles in the processing, will be dwindled into from the prediction piece of prediction generating unit 108 inputs by the portion 2110-2 of dwindling and dwindle the prediction piece, and the piece after dwindling is outputed to subtraction portion 2102 and addition portion 2106.By following treatment S 2405～treatment S 2408, implement to dwindle the Code And Decode processing of difference block, generate decoding and dwindles difference block.Dwindling during piece generate to handle in the decoding of treatment S 2409, is that unit will dwindle the prediction piece and the difference block addition is dwindled in decoding by addition portion 2106 with the pixel, generates decoding and dwindles piece.In the processing and amplifying of treatment S 2414, by enlarging section 2111 piece is dwindled in decoding and zoom into decoding block.
Prediction piece in the treatment S 2509 of Figure 25 dwindles in the processing, will dwindle into from the prediction piece of prediction generating unit 605 inputs and dwindle the prediction piece by dwindling portion 2207, and the piece after dwindling is outputed to addition portion 2203.In treatment S 2505, implement to dwindle the decoding processing of difference block, generate decoding and dwindle difference block.Dwindling during piece generate to handle in the decoding of treatment S 2506, is that unit will dwindle the prediction piece and the difference block addition is dwindled in decoding by addition portion 2203 with the pixel, generates decoding and dwindles piece.In the processing and amplifying of treatment S 2510, by enlarging section 2208 piece is dwindled in decoding and zoom into decoding block.
Can consider following variation in the present embodiment.Following variation is described moving picture encoding device and dynamic image decoding device, yet can implement too for moving picture encoding processing and moving image decoding processing.
(1) Forecasting Methodology of use template matches
In the present embodiment, the template matching method in the prediction generating unit is not limited to the method for Fig. 2.That is, in present embodiment and variation, also can use the prediction signal generation method of the use template matches shown in the 1st execution mode, the 2nd execution mode and their variation.As Figure 18 and shown in Figure 19, can also append selection portion, and prediction generating unit 108,605 can also be replaced as prediction generating unit 1108 shown in Figure 20.And,, thereby will predict that generating unit 108,605 is replaced as prediction generating unit 800 shown in Figure 8 and also can directly uses because the input and output flow process of signal is constant.
So far, as the Forecasting Methodology of using template matches, adopted and used template to come according to the method for reproducing dynamic image data generation forecast piece that is kept in the storage part 107,604.Yet, also have as the intra-prediction of non-patent literature 1 record, generate the method for prediction signal according to the reproducing signal use prescriptive procedure of template.In the method, the coding/decoding that uses the dwindling of present embodiment/processing and amplifying is handled owing to have an effect of cutting down the sign indicating number amount, thereby also is effective.
Figure 36 illustrates the structure of the prediction generating unit in this variation.This figure is the example according to the intra-prediction of the signal generation forecast signal of template.The prediction generating unit 3608 that can be replaced as Figure 36 according to the method for the reproducing signal generation forecast signal of template by the prediction generating unit 108 with Figure 21 and Figure 22 is implemented.In the determination section 3602 of template zone, 13 of use and coded object piece adjacency have been reproduced pixel and have constituted template.In the portion of filling up 3603, use the method shown in the processing 3711 shown in Figure 37, according to the pixel generation forecast piece in the template.In addition, Figure 37 shows 9 kinds of complementing methods, and the present invention can by pre-defined wherein at least a kind implement.And, also can in the portion of filling up, from a plurality of complementing methods, select a kind, and the prediction piece of selected complementing method generation is used in output.At this moment, the system of selection of complementing method does not limit in the present invention.Can transmit the information of selected complementing method, also can be the method for only using encoder (moving picture encoding device) and the shared data of decoder (dynamic image decoding device) to decide.
And shape of template is not limited to shape shown in Figure 37.Be not only the pixel of block boundary, so long as the pixel of reproduction regions of the frame under the coded object piece, even the template that is made of the pixel of leaving block boundary also can realize the present invention.
(2) dwindle/structure of processing and amplifying
In moving picture encoding device shown in Figure 21 2100 and dynamic image decoding device 2200 shown in Figure 22,, generate decoding block by dwindling coded object piece and prediction piece and amplifying the decoded piece that dwindles.Yet, also can realize in different structures even use to dwindle with yard reduction of measuring of processing and amplifying, be not limited to this method.Figure 32 and Figure 34 illustrate other examples of moving picture encoding device, and Figure 33 and Figure 35 illustrate other examples of dynamic image decoding device.
In the moving picture encoding device 3200 of Figure 32, not with the prediction piece, but will dwindle by dwindling portion 3210 by to be unit with the pixel deduct the difference block that the prediction piece generates from the coded object piece.Then, difference block being dwindled in decoded decoding amplified by enlarging section 3211, is that unit will amplify piece and the addition of prediction piece with the pixel, obtains decoding block.
The dynamic image decoding device 3300 of Figure 33 is decoding devices corresponding with moving picture encoding device shown in Figure 32 3200.In the method, not implementing to predict dwindling of piece, amplified by enlarging section 3308 and difference block is dwindled in decoded decoding, is that unit will amplify piece and the addition of prediction piece with the pixel, obtains decoding block.
The moving picture encoding device 3400 of Figure 34 has functional packet with the 2110-2 of the portion that dwindles among Figure 21 and is contained in the structure in the portion filled up in the prediction generating unit 3408.Equally, the dynamic image decoding device 3500 of Figure 35 has the structure in the portion filled up that the portion that dwindles 2207 among Figure 22 is included in the prediction generating unit.At this moment, the portion of filling up can also handle dwindling and fill up to handle and make up, and dwindles the prediction piece so that directly generate.In Fig. 2, Fig. 8 or the portion of filling up 203 shown in Figure 20, filling up the sizes of memory that portion 203 has is to preserve the size of dwindling the prediction piece and can not preserve the prediction piece, directly only obtain needed information according to the dynamic image data of reproduction in the storage part 107, generate and dwindle the prediction piece.In this case, all pixels in the template zone can be as noted above be obtained according to the dynamic image data that reproduces in the storage part 107 in template zone determination section and matching part, can also also implement to dwindle to the template zone, only obtain needed information and implement matching treatment.On the other hand, in dynamic image decoding device shown in Figure 35 3500, template zone determination section also can directly only be obtained needed information according to the dynamic image data of reproduction in the storage part 604.
In addition, no matter the structure of the prediction generating unit in above-mentioned moving picture encoding device 3200,3400 and the dynamic image decoding device 3300,3500 is that structure or structure shown in Figure 36 shown in Figure 2 can both realize.And,, under situation, also can realize, thereby can also be replaced as Fig. 2 and prediction generating unit shown in Figure 20 by appending selection portion with selection portion as Figure 18 and moving picture encoding device 1800 and dynamic image decoding device 1900 shown in Figure 19.And,, thereby Fig. 2 is replaced as prediction generating unit shown in Figure 8 also can directly uses because the input and output flow process of signal is constant.
And, in Figure 21, also can not implement to decode and dwindle the processing and amplifying of piece, but directly be saved in the frame memory.And, in other frame, use and to dwindle the downscaled images that piece obtains by integrated decoding and carry out template matches, generate and dwindle the prediction piece.Then, dwindle coded object piece and the differential coding that dwindles the prediction piece.Perhaps, also can amplify and dwindle the prediction piece, carry out differential coding with the coded object piece.In Figure 22, also can not implement to decode equally and dwindle the processing and amplifying of piece, but directly be saved in the frame memory.In other frame, use by integrated and dwindle the downscaled images that piece obtains and carry out template matches, generate and dwindle the prediction piece.Then, dwindle difference block and dwindle the prediction piece and reproduce and to dwindle piece in the Calais mutually decoded.Perhaps, also can amplify and dwindle the prediction piece, with decoded difference block with amplify after piece reproduce decoding block in the Calais mutually.Like this, even the application process of processing and processing and amplifying is dwindled in change, also can obtain an effect of cutting down the sign indicating number amount.
The method of dwindling processing and processing and amplifying can be a method shown in Figure 23, also can be other example described later.
(3) system of selection of TMP-E, TMP-L
The pattern in Figure 21 and moving picture encoding device 2100 shown in Figure 22 and dynamic image decoding device 2200 expression supposition template zones all is smooth situation, with the structure of the device of the dwindling of piece/processing and amplifying.In fact, in the pattern of image, mix and have flat and characteristic part, thus with the 1st execution mode in the dwindling of the piece/structure of processing and amplifying of not implementing that illustrates make up and implement.Figure 26 illustrates the moving picture encoding device 2600 that Fig. 1 and moving picture encoding device 100,2100 shown in Figure 21 are combined, and Figure 27 illustrates the dynamic image decoding device 2700 that Fig. 6 and dynamic image decoding device 600,2200 shown in Figure 22 are combined.In the drawings, represent identical function with the piece of Fig. 1, Fig. 6, Figure 21, the identical numbering of Figure 22, thereby detailed here.
In Figure 26, the coded object piece that is partitioned into by Region Segmentation portion 101 is input to subtraction portion 102 respectively and is dwindled the 2110-1 of portion.The coded object piece uses 2 kinds of predictive coding methods, and (TMP-E pattern: the coded system of moving picture encoding device 100 shown in Figure 1, TMP-L pattern: the coded system of moving picture encoding device 2100 shown in Figure 21) carry out encoding process, 2 kinds of transform datas (quantization data) are output to switching part 2613 from transformation component 103 and transformation component 2103.And, use 2 kinds of methods to carry out local decoded 2 kinds of decoding blocks and be output to selection portion 2612.In selection portion 2612, from these 2 kinds of decoding blocks, select a kind, this selection information is outputed to switching part 2613 and encoding section 2604.Switching part 2613 outputs to encoding section 2604 to transform data (quantization data) according to selection information.Encoding section will select information and transform data (quantization data) to carry out entropy coding in the lump.
As the system of selection of TMP-E in the selection portion 2612 and TMP-L pattern, for example enumerate following method, that is: the coded object piece that will be imported is separately compared with the difference signal of 2 kinds of decoding blocks, selects the little side of quadratic sum of difference.Yet, in the method, do not converted because sign indicating number is measured, thereby do not selected the high coding method of code efficiency.As the system of selection of considering code efficiency, for example, enumerate the method that non-patent literature 2 is put down in writing.According to this method, the transform data (quantization data) that uses 2 kinds of predictive coding methods to generate is carried out virtual encoder respectively.Then, calculate total distortion (for example quadratic sum of difference) and yard Weighted Guidelines value of measuring between decoding block and the coded object piece respectively, select the little coding method of desired value.So long as this method just can be selected the high predictive coding method of code efficiency.In addition, the coding of transform data (quantization data) can be input to these data selection portion 2612 and be implemented by selection portion 2612, also can be implemented by encoding section 2604, and the information of sign indicating number amount is input to selection portion.
In dynamic image decoding device shown in Figure 27 2700, carry out the entropy decoding by the selection information of 2701 pairs of TMP-E patterns of lsb decoder and TMP-L pattern and the transform data (quantization data) of the selected predictive coding method of use.Switching part 2709 when selecting information representation TMP-E, outputs to inverse transformation portion 602 to transform data (quantization data) according to selection information, when selecting information representation TMP-L pattern, transform data (quantization data) is outputed to inverse transformation portion 2202.Transform data (quantization data) uses selects the coding/decoding method of information representation to decode.
As mentioned above, TMP-E is handled as different predictive coding methods with TMP-L, this selection information is encoded in the coding side.Yet, 2 kinds of predictive coding methods can also be handled as a kind, use the decoded information that coding side and decoding side share (according to reproduced image or belong to the information that the characteristic of reproduced image derives) to select automatically.Can utilize the object signal in template zone as characteristic.For example, the method that has the variance of the object signal of using the template zone.Can consider such method, that is: preestablish threshold value, when variance yields uses TMP-E during greater than threshold value, when variance yields uses TMP-L during less than threshold value.And, except variance yields, also can consider such method, that is: the pixel gradient of the object signal in calculation template zone (difference between adjacent pixels), the pixel count that surpasses threshold value when difference uses TMP-E during more than predetermined number, and the pixel count that surpasses threshold value when difference uses TMP-L when being less than predetermined number.And, can also will compare by the motion vector (can utilize the signal of decoding block to retrieve once more) of the detected motion vector of template matches when the pixel groups in template zone is decoded, under the situation of its difference, use TMP-E, under the situation of its difference, use TMP-L greater than predetermined threshold value less than predetermined threshold value.Can be that benchmark is selected with the size of detected motion vector or the motion vector of adjacent block.Can will use the selection of these variance yields, pixel gradient and motion vector to make up.Under according to the different situation of each side's method selected, can adopt most decisions to carry out final selection, can also realize only under the different situation of method selected, just transmitting the method for selection information, can obtain stable selection result.
In this case, in the dynamic image decoding device 2700 of Figure 27,, thereby be necessary to be provided with separately selection portion because selection information is not decoded.This selection portion is carried out the action identical with the selection portion of code device, and selection information is outputed to switching part 2709.
Figure 26 and the prediction generating unit of respectively installing in 2600,2700 shown in Figure 27 are not limited to structure shown in Figure 2 shown in above-mentioned variation (1).Can also application drawing 8, Figure 20 and structure shown in Figure 36.
In addition, in prediction generating unit 3608 shown in Figure 36, only under the situation of predetermined complementing method, just use such the dwindling of TMP-L/processing and amplifying, in other complementing methods, can also only use the such differential coding of TMP-E.And the prediction piece that can also generate at using selected complementing method by the portion of filling up practices the method that such coding method of TMP-L and TMP-E and self adaptation are selected.
(4) with the self adaptation of existing predictive coding method
A plurality of predictive coding methods that predictive coding method TMP-E shown in above-mentioned and TMP-L can be put down in writing with non-patent literature 1 (outside predictive mode and intra-prediction mode that motion vector is encoded) make up selectivity to use.At this moment, about each predictive coding method, can prepare a plurality of sizes.The optimal selection of predictive coding method and piece size for example can use the method shown in the non-patent literature 2 to realize.By at Figure 26 with shown in Figure 27 respectively install the existing predictive coding method of combination in 2600,2700 (prediction generation method can be carried out the distortion shown in the variation (1)), and expand selection portion and realize.And, can also only make TMP-L and existing predictive coding method self adaptation.In this case, can be at Figure 21 and shown in Figure 22 respectively install 2100,2200, Figure 32 and shown in Figure 33 respectively install 3200,3300 and Figure 34 and any one party of respectively installing in 3400,3500 shown in Figure 35 in the existing predictive coding method of combination.
(5) dwindle/processing and amplifying
The piece that dwindles portion dwindle handle and the enlarging section in the method for piece processing and amplifying be not limited to the method for Figure 23.Figure 28 and Figure 29 illustrate other examples.
In Figure 28, the piece before piece 2801 expressions are reduced, piece is dwindled in piece 2802 expressions.In this example, dwindle piece by generating without the simple pixel sampling of handling 2304 such Filtering Processing.Handle the pixel generation method on the amplification piece in the 2805 expression processing and amplifying.Pixel A～D represents the pixel on the piece 2802, and pixel a～c represents the pixel on the enlarged image.Pixel A～D was owing to be the pixel of dwindling before handling originally, thereby directly was copied into and amplifies piece 2803.About by dwindle handle and by except pixel, as handle pixel a～c of 2305, handle by simple linear interpolation and calculate.The pixel of being represented by square of piece 2803 is handled by the linear interpolation of using adjacent pixels equally and is calculated.In addition, the pixel about the black box of the upper end of amplifying piece 2803 and left end only exists 1 or 2 adjacent pixels of being represented by circle.In this case, utilize the decoded pixel of the adjacent block shown in the pixel groups 2806.In addition, owing to pixel groups 2806 is stored in the storage part, thereby, need from the input path of storage part in installing to the enlarging section in Figure 21, Figure 22, Figure 26, shown in Figure 27 each in order to realize this processing.
In Figure 29, piece 2901 is reduced preceding pieces, and piece is dwindled in piece 2902 expressions, describes dwindling processing method in processing 2904.Dwindle in the method at this, as handle shown in 2904, by utilize with 8 pixels of pixel p adjacency (j, k, l, m, n, o, q, Filtering Processing r) generates the pixel P that dwindles on the piece.Handle the pixel generation method on the amplification piece 2903 in the 2905 expression processing and amplifying.This is handled because 2805 identical with Figure 28, thereby omits explanation.In this case, owing to pixel groups 2906 is stored in the storage part, thereby, in Figure 21, Figure 22, Figure 26, Figure 27, need from the input path of storage part to the enlarging section in order to realize this processing.
In these examples, the size of dwindling piece is that the minification of vertical/horizontal stroke is respectively 1/2, yet minification is not limited thereto.For example, minification can be 1/4, vertical/horizontal minification can be different.
Dwindle/amplification method is not limited to a kind, can selects from a plurality of methods.As system of selection, can the coding side with the frame be unit or with the piece be unit select encoding error absolute value and or the method for quadratic sum minimum, and selection information encoded, can also decision automatically from a plurality of decoding block candidates.As determining method, as long as use from reproduced image or belong to the information that the characteristic of reproduced image derives, for example can consider with the pixel to be the method for unit calculating mean value, perhaps be the method etc. of unit selection median with the pixel.
In addition, the method for dwindling about at the coded object piece does not limit in decoding device.Therefore, as long as it is identical to dwindle the pixel count of piece, just the different methods of dwindling can be applied to predict piece and coded object piece.And in decoding device and decoding processing, regulation coded object piece does not dwindle method.
(6) coding again of amplification piece
In Figure 26 and shown in Figure 27 respectively installing in 2600,2700, by the candidate of the piece after 2111,2208 amplifications of enlarging section as decoding block, yet the candidate of this piece as the prediction piece can also be handled, and select with the prediction block adaptive ground that generates by prediction generating unit 108,605.Piece after being amplified by the enlarging section is because high fdrequency component is limited by Filtering Processing, thereby has by this piece is encoded again and improve the effect of image quality.
Figure 30 and Figure 31 illustrate moving picture encoding device 3000 and the dynamic image decoding device 3100 of implementing this variation respectively.
In the moving picture encoding device 3000 of Figure 30, compare the function of selection portion 3012 and encoding section 3004 and different with Figure 26 from the processing of the transform data (quantization data) of transformation component 2103 output.In the selection portion 2612 of Figure 26, the candidate of 2 kinds of decoding blocks of input, and in the selection portion 3012 of this variation, from the candidate of prediction generating unit 108 and 2 kinds of predictions of enlarging section 2111 inputs piece.Can utilize the method shown in the explanation of Figure 26 as system of selection.Yet, under the situation of the method for utilizing non-patent literature 2,, be necessary the candidate of 2 kinds of prediction pieces is carried out virtual encoder/decoding in order to calculate total distortion and sign indicating number amount.And about TMP-L, the transform data of exporting from transformation component 2103 (quantization data) also is necessary to carry out virtual encoder, and is converted into the sign indicating number amount.Selected prediction piece is output to addition portion 106 and subtraction portion 102, carries out transformed/coded.Select by selection portion under the situation of TMP-L, connect switch 3013, and the transform data (quantization data) from transformation component 2103 outputs is outputed to encoding section 3004.In encoding section 3004, compile from the data of transformation component 103, transformation component 2103 (under the situation of TMP-L) and selection portion (under the necessary situation) and encode.
In addition, implementing under the quantized situation by transformation component 103,2103, when must be higher than the quantization precision of transformation component 103 the quantization precision set of transformation component 2103, the code efficiency raising.Can encode with the sign indicating number amount of lacking than the coded object piece owing to dwindle piece, even thereby improve the quantization precision, the sign indicating number amount does not increase yet.Therefore, has the effect that code efficiency is worsened and can improve image quality.
In the dynamic image decoding device 3001 of Figure 31, compare the function of lsb decoder 3101 and switching part 3109 and different with Figure 27 from the processing of predicting the prediction piece that generating unit 605 is exported.In lsb decoder 3101, at first selection information is carried out the entropy decoding.When the predictive coding method is the TMP-L pattern, the transform data (quantization data) that dwindles piece is carried out the entropy decoding.The transform data (quantization data) that dwindles piece is output to inverse transformation portion 2202 by the control of switching part 3109.The transform data (quantization data) that amplifies piece is carried out the entropy decoding, and output to inverse transformation portion 602 by the control of switching part 3109.And,, be output to the portion of dwindling 2207 by the prediction piece that template matches generates by prediction generating unit 605 by control based on the switch 3110 of selecting information.The difference block that addition portion 603 will obtain from inverse transformation portion 602 and from the enlarging section the 2208 prediction piece additions that obtain generate decoding block.On the other hand, when the predictive coding method is the TMP-E pattern, has carried out the decoded transform data of entropy (quantization data) control by switching part 3109 and be output to inverse transformation portion 602.And,, be output to addition portion 603 by the prediction piece that template matches generates by prediction generating unit 605 by control based on the switch 3110 of selecting information.In addition portion 603, the difference block that will obtain from inverse transformation portion 602 and generate decoding block from the prediction piece addition that prediction generating unit 605 obtains by switch 3110.
In addition, in this variation, can use the mode selecting method shown in the template prediction method shown in the modified embodiment of the present embodiment (1) and variation (3) and (4).
Here, the dynamic image encoding that is used to make moving picture encoding device carry out above-mentioned a series of encoding process is described.As shown in figure 16, dynamic image encoding 1601 is stored in the program storage area 1600a that is formed in the recording medium 1600 that can be read or be equipped with by moving picture encoding device.
Identical by carrying out the function that above-mentioned each module realizes with the function of each inscape of above-mentioned moving picture encoding device 100.That is, cut apart Region Segmentation portion 101, subtraction portion 102, transformation component 103, encoding section 104, inverse transformation portion 105, addition portion 106, the storage part 107 in the moving picture encoding device 100 of function that each module of module 1601b, subtraction block 1601c, conversion module 1601d, coding module 1601e, inverse transform module 1601f, addition module 1601g, memory module 1601h and prediction generation module 1601i realizes and above-mentioned execution mode and predict that the function of generating unit 108 is identical by execution area.And, by carrying out template zone decision module 1601j, matching module 1601k and filling up template regional determination section 201, the matching part 202 in the moving picture encoding device 100 of function that each module of module 1601m realizes and above-mentioned execution mode and fill up the function of portion 203 identical.
Next, the moving image decoding program that is used to make dynamic image decoding device carry out above-mentioned a series of decoding processing is described.As shown in figure 17, moving image decoding program 1701 is stored in the program storage area 1700a that is formed in the recording medium 1700 that can be read or be equipped with by dynamic image decoding device.
Moving image decoding program 1701 constitutes to be had: moving image decoding is handled primary module 1701a, decoder module 1701b, inverse transform module 1701c, addition module 1701d, memory module 1701e and the prediction generation module 1701f that unifies to control.And prediction generation module 1701f has template zone decision module 1701g, matching module 1701h and fills up module 1701i.
Identical by carrying out the function that above-mentioned each module realizes with the function of each inscape of above-mentioned dynamic image decoding device 600.That is lsb decoder 601, inverse transformation portion 602, addition portion 603, the storage part 604 in the dynamic image decoding device 600 of the function that realizes by each module of carrying out decoder module 1701b, inverse transform module 1701c, addition module 1701d, memory module 1701e and prediction generation module 1701f and above-mentioned execution mode and predict that the function of generating unit 605 is identical.And, by carrying out template zone decision module 1701g, matching module 1701h and filling up moving picture encoding device 100 or the template in the dynamic image decoding device 600 regional determination section 201, the matching part 202 of function that each module of module 1701i realizes and above-mentioned execution mode and fill up the function of portion 203 identical.
In addition, dynamic image encoding 1601 and moving image decoding program 1701 can adopt such structure, that is: its part or all transmit via transmission mediums such as communication lines, and receive and record (comprising installation) by another equipment.
Here, dynamic image encoding 1601 and the moving image decoding program 1701 corresponding with the moving picture encoding device 100 of the Fig. 1 shown in the 1st execution mode and Fig. 6 record and dynamic image decoding device 600 are described, yet variation, the 2nd and the 3rd execution mode and variation thereof about the 1st execution mode, implement the module of the function of moving picture encoding device and dynamic image decoding device by preparation, can constitute dynamic image encoding and moving image decoding program, thereby be also contained in the present invention.
Claims (34)
1. moving picture encoding device, it is that unit encodes to dynamic image data with the piece, it is characterized in that this moving picture encoding device has:
Cutting unit, its two field picture that will constitute above-mentioned dynamic image data is divided into a plurality of coded object pieces;
Coding unit, it is encoded to above-mentioned coded object piece;
The reproduced image generation unit, its reproducing signal that generates above-mentioned coded object piece is a decoding block;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
The prediction signal generation unit, its use according to the position relation of regulation and above-mentioned coded object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data by the said memory cells storage generates, the prediction signal that generates above-mentioned coded object piece is promptly predicted piece;
To be unit with the pixel deduct the difference signal that above-mentioned prediction piece generates this coded object piece from above-mentioned coded object piece to above-mentioned coding unit is difference block, and this difference block is encoded;
The reproducing signal that above-mentioned reproduced image generation unit generates by the difference block behind the above-mentioned coding unit coding is the decoded difference value piece, and is that unit is with this decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
2. moving picture encoding device, it is that unit encodes to dynamic image data with the piece, it is characterized in that this moving picture encoding device has:
Cutting unit, its two field picture that will constitute above-mentioned dynamic image data is divided into a plurality of coded object pieces;
Coding unit, it is encoded to above-mentioned coded object piece;
The reproduced image generation unit, its reproducing signal that generates above-mentioned coded object piece is a decoding block;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal;
Retrieval unit, it is from the dynamic image data of reproduction by the said memory cells storage, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal concerns with above-mentioned coded object piece adjacency with the position of stipulating and belongs to the dynamic image data of being stored by said memory cells of reproduction; And
Prediction signal decision unit, it determines the prediction signal of above-mentioned coded object piece promptly to predict piece based on the position relation of pixel groups that is retrieved by above-mentioned retrieval unit and afore mentioned rules according to the dynamic image data that reproduces by the said memory cells storage;
To be unit with the pixel deduct the difference signal that above-mentioned prediction piece generates this coded object piece from above-mentioned coded object piece to above-mentioned coding unit is difference block, and this difference block is encoded;
The reproducing signal that above-mentioned reproduced image generation unit generates by the difference block behind the above-mentioned coding unit coding is the decoded difference value piece, and is that unit is with this decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
3. moving picture encoding device according to claim 2 is characterized in that, this moving picture encoding device also has:
Estimation unit, it compares with above-mentioned template with by the dynamic image data of reproduction of said memory cells storage, estimates the spatial continuity of the image of above-mentioned coded object piece according to this comparative result; And
Setup unit, it further cuts apart above-mentioned coded object piece according to the spatial continuity of the image that is estimated by above-mentioned estimation unit, and the coded object piece after this is cut apart is set at new coded object piece, and sets the template at this new coded object piece.
4. dynamic image decoding device, to be unit with the piece with the coded data of dynamic image data be reproduced as reproduces dynamic image data for it, it is characterized in that this dynamic image decoding device has:
Decoding unit, its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded;
The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated above-mentioned decoder object piece by the coded data after the above-mentioned decoding unit decodes;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
The prediction signal generation unit, its use according to the position relation of regulation and above-mentioned decoder object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data by the said memory cells storage generates, the prediction signal that generates this decoder object piece is promptly predicted piece;
The difference signal that above-mentioned decoding unit generates above-mentioned decoder object piece is the decoded difference value piece;
Above-mentioned reproduced image generation unit is that unit is with above-mentioned decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
5. dynamic image decoding device, to be unit with the piece with the coded data of dynamic image data be reproduced as reproduces dynamic image data for it, it is characterized in that this dynamic image decoding device has:
Decoding unit, its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded;
The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated above-mentioned decoder object piece by the coded data after the above-mentioned decoding unit decodes;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal;
Retrieval unit, it is from the dynamic image data of reproduction by the said memory cells storage, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal concerns with above-mentioned decoder object piece adjacency with the position of stipulating and belongs to the dynamic image data of being stored by said memory cells of reproduction; And
Prediction signal decision unit, it determines the prediction signal of above-mentioned decoder object piece promptly to predict piece based on the position relation of pixel groups that is retrieved by above-mentioned retrieval unit and afore mentioned rules according to the dynamic image data that reproduces by the said memory cells storage;
The difference signal that above-mentioned decoding unit generates above-mentioned decoder object piece is the decoded difference value piece;
Above-mentioned reproduced image generation unit is that unit is with above-mentioned decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
6. dynamic image decoding device according to claim 5 is characterized in that, this dynamic image decoding device also has:
Estimation unit, it compares with above-mentioned template with by the dynamic image data of reproduction of said memory cells storage, estimates the spatial continuity of the image of above-mentioned decoder object piece according to this comparative result; And
Setup unit, it is according to the spatial continuity of the image that is estimated by above-mentioned estimation unit, further cut apart above-mentioned decoder object piece, the above-mentioned decoder object piece after this is cut apart is set at new decoder object piece, and sets the template at this new decoder object piece.
7. according to any described moving picture encoding device in the claim 1～3, it is characterized in that,
Above-mentioned coding unit makes the processing of dwindling that pixel count reduces by utilizing prescriptive procedure, above-mentioned difference block is generated as pixel count dwindles difference block than what this difference block was lacked, and this is dwindled difference block encode;
The above-mentioned reproducing signal that dwindles difference block of above-mentioned reproduced image generation unit generation is promptly decoded and is dwindled difference block, by the processing and amplifying of utilizing prescriptive procedure that pixel count is increased, dwindles difference block generating solution code block according to this decoding.
8. according to any described dynamic image decoding device in the claim 4～6, it is characterized in that,
Above-mentioned decoding unit is by decoding to above-mentioned coded data, and difference block is dwindled in the decoding that the generation pixel count lacks than above-mentioned difference block;
The processing and amplifying of above-mentioned reproduced image generation unit by utilizing the afore mentioned rules method that pixel count is increased dwindled difference block generating solution code block according to above-mentioned decoding.
9. moving picture encoding device, it is that unit encodes to dynamic image data with the piece, it is characterized in that this moving picture encoding device has:
Cutting unit, its two field picture that will constitute above-mentioned dynamic image data are divided into a plurality of coded object pieces as the zone that becomes above-mentioned coded object;
Coding unit, it is encoded to above-mentioned coded object piece;
The reproduced image generation unit, its reproducing signal that generates above-mentioned coded object piece is a decoding block;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
The prediction signal generation unit, its basis concerns with above-mentioned coded object piece adjacency with the position of stipulating and belongs to the reproducing signal of being stored by said memory cells that reproduces dynamic image data that the method that use is scheduled to generates the prediction signal of above-mentioned coded object piece and promptly predicts piece;
Above-mentioned coding unit is a difference block at the difference signal that to be unit with the pixel deduct this coded object piece behind the above-mentioned prediction piece from above-mentioned coded object piece, by utilizing prescriptive procedure to make the processing of dwindling of pixel count minimizing, generate pixel count and dwindle difference block, and this is dwindled difference block encode than what this difference block was lacked;
The above-mentioned reproducing signal that dwindles difference block of above-mentioned reproduced image generation unit generation is promptly decoded and is dwindled difference block, by the processing and amplifying of utilizing prescriptive procedure that pixel count is increased, dwindles difference block generating solution code block according to this decoding.
10. dynamic image decoding device, to be unit with the piece with the coded data of dynamic image data be reproduced as reproduces dynamic image data for it, it is characterized in that this dynamic image decoding device has:
Decoding unit, its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded;
The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated above-mentioned decoder object piece by the coded data after the above-mentioned decoding unit decodes;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
The prediction signal generation unit, its basis concerns with above-mentioned decoder object piece adjacency with the position of stipulating and belongs to the reproducing signal of being stored by said memory cells that reproduces dynamic image data that the method that use is scheduled to generates the prediction signal of above-mentioned decoder object piece and promptly predicts piece;
Above-mentioned decoding unit is by decoding to above-mentioned coded data, and generating pixel count is that difference block is dwindled in the decoding that difference block is lacked than the difference signal of above-mentioned decoder object piece;
The processing and amplifying of above-mentioned reproduced image generation unit by utilizing prescriptive procedure that pixel count is increased dwindled difference block generating solution code block according to above-mentioned decoding.
11. according to claim 7 or 9 described moving picture encoding devices, it is characterized in that,
Above-mentioned coding unit above-mentioned dwindle to handle be applied to above-mentioned coded object piece and above-mentioned prediction piece and obtain dwindling piece respectively and dwindle the prediction piece, to be unit with the pixel dwindle piece from this deducts this and dwindles the prediction piece and generate and dwindle difference block;
The reproducing signal that dwindles difference block after above-mentioned reproduced image generation unit generation is encoded by above-mentioned coding unit is promptly decoded and is dwindled difference block, to be unit with the pixel dwindle difference block with this decoding and above-mentionedly dwindle the addition of prediction piece and generate decoding and dwindle piece, and processing and amplifying is applied to this decoding dwindles piece and generate decoding block.
12. according to claim 7 or 9 described moving picture encoding devices, it is characterized in that,
Above-mentioned coding unit by above-mentioned dwindle to handle be applied to above-mentioned difference block and generate and dwindle difference block;
The reproducing signal that dwindles difference block after above-mentioned reproduced image generation unit generation is encoded by above-mentioned coding unit is promptly decoded and is dwindled piece, dwindle piece and generate the decoded difference value piece by processing and amplifying being applied to this decoding, and be that unit generates decoding block with this decoded difference value piece and the addition of above-mentioned prediction piece with the pixel.
13. according to Claim 8 or 10 described dynamic image decoding devices, it is characterized in that, above-mentioned reproduced image generation unit with above-mentioned dwindle to handle be applied to above-mentioned prediction piece and obtain dwindling the prediction piece, to be unit with the pixel dwindle difference block and this with above-mentioned decoding dwindles the addition of prediction piece and generates decoding and dwindle piece, and processing and amplifying is applied to this decoding dwindles piece and generate decoding block.
14. according to Claim 8 or 10 described dynamic image decoding devices, it is characterized in that, above-mentioned reproduced image generation unit dwindles difference block and generates the decoded difference value piece by above-mentioned processing and amplifying being applied to above-mentioned decoding, and is that unit generates decoding block with this decoded difference value piece and the addition of above-mentioned prediction piece with the pixel.
15. any described moving picture encoding device according in the claim 2,3,7,11,12 is characterized in that, 1 template is selected in above-mentioned prediction signal decision unit from variform a plurality of above-mentioned templates.
16. moving picture encoding device according to claim 15, it is characterized in that, above-mentioned 1 template is selected with reference to by the reproducing signal that reproduces dynamic image data of said memory cells storage or the information relevant with this reproducing signal in above-mentioned prediction signal decision unit.
17. moving picture encoding device according to claim 15 is characterized in that, above-mentioned coding unit is encoded to the information of specifying the template of being selected by above-mentioned prediction signal decision unit.
18. any described dynamic image decoding device according in the claim 5,6,8,13,14 is characterized in that, 1 template is selected in above-mentioned prediction signal decision unit from variform a plurality of above-mentioned templates.
19. dynamic image decoding device according to claim 18, it is characterized in that, above-mentioned 1 template is selected with reference to by the reproducing signal that reproduces dynamic image data of said memory cells storage or the information relevant with this reproducing signal in above-mentioned prediction signal decision unit.
20. dynamic image decoding device according to claim 18 is characterized in that,
Above-mentioned decoding unit is decoded to the information of specifying selected template;
1 template is selected with reference to the information by the selected template of appointment after the above-mentioned decoding unit decodes in above-mentioned prediction signal decision unit from variform a plurality of above-mentioned templates.
21. the motion image encoding method in a motion image encoding method, it is moving picture encoding device that to be unit with the piece encode to dynamic image data is characterized in that this motion image encoding method has:
Segmentation procedure, its two field picture that will constitute above-mentioned dynamic image data is divided into a plurality of coded object pieces;
Coding step, it is encoded to above-mentioned coded object piece;
Reproduced image generates step, and its reproducing signal that generates above-mentioned coded object piece is a decoding block;
Storing step, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
Prediction signal generates step, its use according to the position relation of regulation and above-mentioned coded object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data stored generates in above-mentioned storing step, the prediction signal that generates above-mentioned coded object piece is promptly predicted piece;
In above-mentioned coding step, to be unit with the pixel deduct the difference signal that above-mentioned prediction piece generates this coded object piece from above-mentioned coded object piece is difference block, and this difference block is encoded;
Generate in the step at above-mentioned reproduced image, the reproducing signal that is created on the difference block behind the coding in the above-mentioned coding step is the decoded difference value piece, and is that unit is with this decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
22. the motion image encoding method in a motion image encoding method, it is moving picture encoding device that to be unit with the piece encode to dynamic image data is characterized in that this motion image encoding method has:
Segmentation procedure, its two field picture that will constitute above-mentioned dynamic image data is divided into a plurality of coded object pieces;
Coding step, it is encoded to above-mentioned coded object piece;
Reproduced image generates step, and its reproducing signal that generates above-mentioned coded object piece is a decoding block;
Storing step, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal;
Searching step, it is from the dynamic image data of storing above-mentioned storing step of reproduction, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal concerns with above-mentioned coded object piece adjacency with the position of stipulating and belong to the dynamic image data of storing of reproduction in above-mentioned storing step; And
The prediction signal deciding step, it determines the prediction signal of above-mentioned coded object piece promptly to predict piece based on the position relation of pixel groups that retrieves in above-mentioned searching step and afore mentioned rules according to the dynamic image data of storing in above-mentioned storing step that reproduces;
In above-mentioned coding step, to be unit with the pixel deduct the difference signal that above-mentioned prediction piece generates this coded object piece from above-mentioned coded object piece is difference block, and this difference block is encoded;
Generate in the step at above-mentioned reproduced image, the reproducing signal that is created on the difference block behind the coding in the above-mentioned coding step is the decoded difference value piece, and is that unit is with this decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
23. dynamic image decoding method, it is, and to be unit with the piece be reproduced as dynamic image decoding method in the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data, it is characterized in that this dynamic image decoding method has:
Decoding step, its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded;
Reproduced image generates step, and it is a decoding block according to the reproducing signal that decoded coded data in above-mentioned decoding step generates above-mentioned decoder object piece;
Storing step, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
Prediction signal generates step, its use according to the position relation of regulation and above-mentioned decoder object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data stored generates in above-mentioned storing step, the prediction signal that generates this decoder object piece is promptly predicted piece;
In above-mentioned decoding step, the difference signal that generates above-mentioned decoder object piece is the decoded difference value piece;
Generating in the step at above-mentioned reproduced image, is that unit is with above-mentioned decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
24. dynamic image decoding method, it is, and to be unit with the piece be reproduced as dynamic image decoding method in the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data, it is characterized in that this dynamic image decoding method has:
Decoding step, its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded;
Reproduced image generates step, and it is a decoding block according to the reproducing signal that decoded coded data in above-mentioned decoding step generates above-mentioned decoder object piece;
Storing step, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal;
Searching step, it is from the dynamic image data of storing above-mentioned storing step of reproduction, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal concerns with above-mentioned decoder object piece adjacency with the position of stipulating and belong to the dynamic image data of storing of reproduction in above-mentioned storing step; And
The prediction signal deciding step, it determines the prediction signal of above-mentioned decoder object piece promptly to predict piece based on the position relation of pixel groups that retrieves in above-mentioned searching step and afore mentioned rules according to the dynamic image data of storing in above-mentioned storing step that reproduces;
In above-mentioned decoding step, the difference signal that generates above-mentioned decoder object piece is the decoded difference value piece;
Generating in the step at above-mentioned reproduced image, is that unit is with above-mentioned decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
25. the motion image encoding method in a motion image encoding method, it is moving picture encoding device that to be unit with the piece encode to dynamic image data is characterized in that this motion image encoding method has:
Segmentation procedure, its two field picture that will constitute above-mentioned dynamic image data are divided into a plurality of coded object pieces as the zone that becomes above-mentioned coded object;
Coding step, it is encoded to above-mentioned coded object piece;
Reproduced image generates step, and its reproducing signal that generates above-mentioned coded object piece is a decoding block;
Storing step, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
Prediction signal generates step, it concerns with above-mentioned coded object piece adjacency according to the position with regulation and belongs to the reproducing signal of storing that reproduces dynamic image data that in above-mentioned storing step the prediction signal of using predetermined method to generate above-mentioned coded object piece is promptly predicted piece;
In above-mentioned coding step, at the difference signal that to be unit with the pixel deduct this coded object piece behind the above-mentioned prediction piece from above-mentioned coded object piece is difference block, by utilizing prescriptive procedure to make the processing of dwindling of pixel count minimizing, generate pixel count and dwindle difference block, and this is dwindled difference block encode than what this difference block was lacked;
Generate in the step at above-mentioned reproduced image, generates the above-mentioned reproducing signal that dwindles difference block and promptly decode and dwindle difference block, and, dwindle difference block generating solution code block according to this decoding by utilizing prescriptive procedure to make the processing and amplifying of pixel count increase.
26. dynamic image decoding method, it is, and to be unit with the piece be reproduced as dynamic image decoding method in the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data, it is characterized in that this dynamic image decoding method has:
Decoding step, its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded;
Reproduced image generates step, and it is a decoding block according to the reproducing signal that decoded coded data in above-mentioned decoding step generates above-mentioned decoder object piece;
Storing step, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
Prediction signal generates step, it concerns with above-mentioned decoder object piece adjacency according to the position with regulation and belongs to the reproducing signal of storing that reproduces dynamic image data that in above-mentioned storing step the prediction signal of using predetermined method to generate above-mentioned decoder object piece is promptly predicted piece;
In above-mentioned decoding step, by above-mentioned coded data is decoded, the generation pixel count is that difference block is dwindled in the decoding that difference block is lacked than the difference signal of above-mentioned decoder object piece;
Generate in the step at above-mentioned reproduced image,, dwindle difference block generating solution code block according to above-mentioned decoding by the processing and amplifying of utilizing prescriptive procedure that pixel count is increased.
27. a dynamic image encoding, its control is the moving picture encoding device that unit encodes to dynamic image data with the piece, it is characterized in that, this dynamic image encoding makes above-mentioned moving picture encoding device as playing a role with lower unit:
Cutting unit, its two field picture that will constitute above-mentioned dynamic image data is divided into a plurality of coded object pieces;
Coding unit, it is encoded to above-mentioned coded object piece;
The reproduced image generation unit, its reproducing signal that generates above-mentioned coded object piece is a decoding block;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
The prediction signal generation unit, its use according to the position relation of regulation and above-mentioned coded object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data by the said memory cells storage generates, the prediction signal that generates above-mentioned coded object piece is promptly predicted piece;
To be unit with the pixel deduct the difference signal that above-mentioned prediction piece generates this coded object piece from above-mentioned coded object piece to above-mentioned coding unit is difference block, and this difference block is encoded;
The reproducing signal that above-mentioned reproduced image generation unit generates by the difference block behind the above-mentioned coding unit coding is the decoded difference value piece, and is that unit is with this decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
28. a dynamic image encoding, its control is the moving picture encoding device that unit encodes to dynamic image data with the piece, it is characterized in that, this dynamic image encoding makes above-mentioned moving picture encoding device as playing a role with lower unit:
Cutting unit, its two field picture that will constitute above-mentioned dynamic image data is divided into a plurality of coded object pieces;
Coding unit, it is encoded to above-mentioned coded object piece;
The reproduced image generation unit, its reproducing signal that generates above-mentioned coded object piece is a decoding block;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal;
Retrieval unit, it is from the dynamic image data of reproduction by the said memory cells storage, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal concerns with above-mentioned coded object piece adjacency with the position of stipulating and belongs to the dynamic image data of being stored by said memory cells of reproduction; And
Prediction signal decision unit, it determines the prediction signal of above-mentioned coded object piece promptly to predict piece based on the position relation of pixel groups that is retrieved by above-mentioned retrieval unit and afore mentioned rules according to the dynamic image data that reproduces by the said memory cells storage;
To be unit with the pixel deduct the difference signal that above-mentioned prediction piece generates this coded object piece from above-mentioned coded object piece to above-mentioned coding unit is difference block, and this difference block is encoded;
The reproducing signal that above-mentioned reproduced image generation unit generates by the difference block behind the above-mentioned coding unit coding is the decoded difference value piece, and is that unit is with this decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
29. moving image decoding program, to be unit with the piece be reproduced as the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data in its control, it is characterized in that this moving image decoding program makes above-mentioned dynamic image decoding device as playing a role with lower unit:
Decoding unit, its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded;
The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated above-mentioned decoder object piece by the coded data after the above-mentioned decoding unit decodes;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
The prediction signal generation unit, its use according to the position relation of regulation and above-mentioned decoder object piece in abutting connection with and belong to the template that the reproducing signal that reproduces dynamic image data by the said memory cells storage generates, the prediction signal that generates this decoder object piece is promptly predicted piece;
The difference signal that above-mentioned decoding unit generates above-mentioned decoder object piece is the decoded difference value piece;
Above-mentioned reproduced image generation unit is that unit is with above-mentioned decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
30. moving image decoding program, to be unit with the piece be reproduced as the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data in its control, it is characterized in that this moving image decoding program makes above-mentioned dynamic image decoding device as playing a role with lower unit:
Decoding unit, its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded;
The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated above-mentioned decoder object piece by the coded data after the above-mentioned decoding unit decodes;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal;
Retrieval unit, it is from the dynamic image data of reproduction by the said memory cells storage, retrieval and the high pixel groups of template correlation that generates according to reproducing signal, wherein, above-mentioned reproducing signal concerns with above-mentioned decoder object piece adjacency with the position of stipulating and belongs to the dynamic image data of being stored by said memory cells of reproduction; And
Prediction signal decision unit, it determines the prediction signal of above-mentioned decoder object piece promptly to predict piece based on the position relation of pixel groups that is retrieved by above-mentioned retrieval unit and afore mentioned rules according to the dynamic image data that reproduces by the said memory cells storage;
The difference signal that above-mentioned decoding unit generates above-mentioned decoder object piece is the decoded difference value piece;
Above-mentioned reproduced image generation unit is that unit is with above-mentioned decoded difference value piece and the addition of above-mentioned prediction piece, generating solution code block with the pixel.
31. a dynamic image encoding, its control is the moving picture encoding device that unit encodes to dynamic image data with the piece, it is characterized in that, this dynamic image encoding makes above-mentioned moving picture encoding device as playing a role with lower unit:
Cutting unit, its two field picture that will constitute above-mentioned dynamic image data are divided into a plurality of coded object pieces as the zone that becomes above-mentioned coded object;
Coding unit, it is encoded to above-mentioned coded object piece;
The reproduced image generation unit, its reproducing signal that generates above-mentioned coded object piece is a decoding block;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
The prediction signal generation unit, its basis concerns with above-mentioned coded object piece adjacency with the position of stipulating and belongs to the reproducing signal of being stored by said memory cells that reproduces dynamic image data that the method that use is scheduled to generates the prediction signal of above-mentioned coded object piece and promptly predicts piece;
Above-mentioned coding unit is a difference block at the difference signal that to be unit with the pixel deduct this coded object piece behind the above-mentioned prediction piece from above-mentioned coded object piece, by utilizing prescriptive procedure to make the processing of dwindling of pixel count minimizing, generate pixel count and dwindle difference block, and this is dwindled difference block encode than what this difference block was lacked;
The above-mentioned reproducing signal that dwindles difference block of above-mentioned reproduced image generation unit generation is promptly decoded and is dwindled difference block, and the processing and amplifying by utilizing prescriptive procedure that pixel count is increased, and dwindles difference block generating solution code block according to this decoding.
32. moving image decoding program, to be unit with the piece be reproduced as the dynamic image decoding device that reproduces dynamic image data with the coded data of dynamic image data in its control, it is characterized in that this moving image decoding program makes above-mentioned dynamic image decoding device as playing a role with lower unit:
Decoding unit, its needed coded data of reproduction to the decoder object piece that becomes decoder object is decoded;
The reproduced image generation unit, it is a decoding block according to the reproducing signal that is generated above-mentioned decoder object piece by the coded data after the above-mentioned decoding unit decodes;
Memory cell, the dynamic image data of reproduction that its storage generates according to above-mentioned reproducing signal; And
The prediction signal generation unit, its basis concerns with above-mentioned decoder object piece adjacency with the position of stipulating and belongs to the reproducing signal of being stored by said memory cells that reproduces dynamic image data that the method that use is scheduled to generates the prediction signal of above-mentioned decoder object piece and promptly predicts piece;
Above-mentioned decoding unit is by decoding to above-mentioned coded data, and the generation pixel count is that difference block is dwindled in the decoding that difference block is lacked than the difference signal of above-mentioned decoder object piece;
The processing and amplifying of above-mentioned reproduced image generation unit by utilizing prescriptive procedure that pixel count is increased dwindled difference block generating solution code block according to above-mentioned decoding.
33. a moving picture encoding device, it is encoded to dynamic image data, it is characterized in that, this moving picture encoding device has:
Cutting unit, its two field picture that will constitute above-mentioned dynamic image data are divided into a plurality of zones as the zone that becomes above-mentioned coded object;
Coding unit, it is encoded to each the regional image after being cut apart by above-mentioned cutting unit;
The reproduced image generation unit, it generates the reproduced image by the image behind the above-mentioned coding unit coding;
Memory cell, the reproduced image that its storage is generated by above-mentioned reproduced image generation unit;
Retrieval unit, its from reproduced image by said memory cells storage, retrieval with the zone of the position relation of regulation and the image of the coded object that becomes above-mentioned coding unit in abutting connection with and with zone as the high image of the reproduced image correlation in the template zone of the part of the reproduced image of storing by said memory cells; And
Prediction signal decision unit, it is based on position relation of zone that is retrieved by above-mentioned retrieval unit and afore mentioned rules, according to the prediction signal by the zone of the above-mentioned coded object of reproduced image decision becoming of said memory cells storage;
Above-mentioned coding unit generates by the prediction signal of above-mentioned prediction signal decision unit decision and becomes difference signal between the image in zone of above-mentioned coded object, and this difference signal is encoded.
34. a dynamic image decoding device, the dynamic image data after it is encoded to the two field picture that is divided into a plurality of zones is decoded, and it is characterized in that this dynamic image decoding device has:
Decoding unit, its above-mentioned each regional data after to above-mentioned coding are decoded;
The reproduced image generation unit, it is according to generating reproduced image by the image after the above-mentioned decoding unit decodes;
Memory cell, the image that its storage is generated by above-mentioned reproduced image generation unit;
Retrieval unit, its from reproduced image by said memory cells storage, retrieval with the zone of the position relation of regulation and the image of the decoder object that becomes above-mentioned decoding unit in abutting connection with and with zone as the high image of the reproduced image correlation in the template zone of the part of the reproduced image of storing by said memory cells; And
Prediction signal decision unit, it is based on position relation of zone that is retrieved by above-mentioned retrieval unit and afore mentioned rules, according to the prediction signal by the zone of the above-mentioned decoder object of reproduced image decision becoming of said memory cells storage;
Above-mentioned reproduced image generation unit is by generating by the prediction signal of above-mentioned prediction signal decision unit decision with by obtaining reproduced image with signal between the image after the above-mentioned decoding unit decodes.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP196351/2005 | 2005-07-05 | ||
| JP2005196351 | 2005-07-05 | ||
| JP094391/2006 | 2006-03-30 |
Related Child Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210539613.3A Division CN103024382B (en) | 2005-07-05 | 2006-07-05 | Dynamic image encoding device, dynamic image encoding method, dynamic image decoding device and dynamic image decoding method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101218829A true CN101218829A (en) | 2008-07-09 |
Family
ID=39624323
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2006800246255A Pending CN101218829A (en) | 2005-07-05 | 2006-07-05 | Dynamic image encoding device, dynamic image encoding method, dynamic image encoding program, dynamic image decoding device, dynamic image decoding method, and dynamic image decoding program |
Country Status (2)
| Country | Link |
|---|---|
| JP (2) | JP5394463B2 (en) |
| CN (1) | CN101218829A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102106149A (en) * | 2008-08-08 | 2011-06-22 | 夏普株式会社 | Dynamic image encoding device and dynamic image decoding device |
| CN102160381A (en) * | 2008-09-24 | 2011-08-17 | 索尼公司 | Image processing device and method |
| CN102160379A (en) * | 2008-09-24 | 2011-08-17 | 索尼公司 | Image processing apparatus and image processing method |
| CN102160382A (en) * | 2008-09-24 | 2011-08-17 | 索尼公司 | Image processing device and method |
| CN102160384A (en) * | 2008-09-24 | 2011-08-17 | 索尼公司 | Image processing device and method |
| CN102318347A (en) * | 2009-02-20 | 2012-01-11 | 索尼公司 | Image processing device and method |
| CN102318346A (en) * | 2009-02-20 | 2012-01-11 | 索尼公司 | Image processing device and method |
| CN102422643A (en) * | 2009-05-15 | 2012-04-18 | 索尼公司 | Image processing device, method, and program |
| CN102640496A (en) * | 2009-11-26 | 2012-08-15 | Jvc建伍株式会社 | Image encoding apparatus, image decoding apparatus, image encoding method, and image decoding method |
| CN102833533A (en) * | 2011-06-15 | 2012-12-19 | 富士通株式会社 | Video decoding apparatus, video coding apparatus, video decoding method, video coding method, and storage medium |
| CN103141102A (en) * | 2010-09-30 | 2013-06-05 | 富士通株式会社 | Image decoding method, image encoding method, image decoding device, image encoding device, image decoding program, and image encoding program |
| CN104104961A (en) * | 2013-04-10 | 2014-10-15 | 华为技术有限公司 | Video encoding method, decoding method and apparatus |
| CN106878733A (en) * | 2012-09-24 | 2017-06-20 | 株式会社Ntt都科摩 | Dynamic image prediction decoding device and dynamic image prediction decoding method |
| CN113596475A (en) * | 2021-06-24 | 2021-11-02 | 浙江大华技术股份有限公司 | Image/video encoding method, apparatus, system, and computer-readable storage medium |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3364655A4 (en) * | 2015-11-11 | 2018-10-10 | Samsung Electronics Co., Ltd. | Method and apparatus for decoding video, and method and apparatus for encoding video |
| CN114222131B (en) * | 2021-12-16 | 2023-10-27 | 北京数码视讯技术有限公司 | Low-delay video coding method and device |
| EP4346200A1 (en) * | 2022-09-27 | 2024-04-03 | Beijing Xiaomi Mobile Software Co., Ltd. | Encoding/decoding video picture data |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3032088B2 (en) * | 1992-03-03 | 2000-04-10 | 株式会社東芝 | Video encoding device |
| JPH0795566A (en) * | 1993-09-21 | 1995-04-07 | Nippon Telegr & Teleph Corp <Ntt> | Method and device for encoding image |
| JP3343554B1 (en) * | 1996-05-28 | 2002-11-11 | 松下電器産業株式会社 | Image prediction decoding method and image prediction encoding device |
| JP3466032B2 (en) * | 1996-10-24 | 2003-11-10 | 富士通株式会社 | Video encoding device and decoding device |
| US6289052B1 (en) * | 1999-06-07 | 2001-09-11 | Lucent Technologies Inc. | Methods and apparatus for motion estimation using causal templates |
| JP3936335B2 (en) * | 2001-11-30 | 2007-06-27 | 株式会社エヌ・ティ・ティ・ドコモ | Moving picture encoding apparatus, moving picture decoding apparatus, moving picture encoding method, moving picture decoding method, program, and computer-readable recording medium storing program |
-
2006
- 2006-07-05 CN CNA2006800246255A patent/CN101218829A/en active Pending
-
2011
- 2011-10-24 JP JP2011233205A patent/JP5394463B2/en active Active
-
2013
- 2013-09-04 JP JP2013183151A patent/JP5650294B2/en active Active
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102106149A (en) * | 2008-08-08 | 2011-06-22 | 夏普株式会社 | Dynamic image encoding device and dynamic image decoding device |
| CN102160381A (en) * | 2008-09-24 | 2011-08-17 | 索尼公司 | Image processing device and method |
| CN102160379A (en) * | 2008-09-24 | 2011-08-17 | 索尼公司 | Image processing apparatus and image processing method |
| CN102160382A (en) * | 2008-09-24 | 2011-08-17 | 索尼公司 | Image processing device and method |
| CN102160384A (en) * | 2008-09-24 | 2011-08-17 | 索尼公司 | Image processing device and method |
| CN102318347B (en) * | 2009-02-20 | 2014-03-12 | 索尼公司 | Image processing device and method |
| CN102318347A (en) * | 2009-02-20 | 2012-01-11 | 索尼公司 | Image processing device and method |
| CN102318346A (en) * | 2009-02-20 | 2012-01-11 | 索尼公司 | Image processing device and method |
| CN102422643A (en) * | 2009-05-15 | 2012-04-18 | 索尼公司 | Image processing device, method, and program |
| CN102640496B (en) * | 2009-11-26 | 2015-08-19 | Jvc建伍株式会社 | Picture coding device, picture decoding apparatus, method for encoding images and picture decoding method |
| CN102640496A (en) * | 2009-11-26 | 2012-08-15 | Jvc建伍株式会社 | Image encoding apparatus, image decoding apparatus, image encoding method, and image decoding method |
| CN103141102A (en) * | 2010-09-30 | 2013-06-05 | 富士通株式会社 | Image decoding method, image encoding method, image decoding device, image encoding device, image decoding program, and image encoding program |
| CN103141102B (en) * | 2010-09-30 | 2016-07-13 | 富士通株式会社 | Image decryption method, image encryption method, image decrypting device, image encrypting apparatus, image decryption program and image encryption program |
| CN102833533A (en) * | 2011-06-15 | 2012-12-19 | 富士通株式会社 | Video decoding apparatus, video coding apparatus, video decoding method, video coding method, and storage medium |
| CN104539963A (en) * | 2011-06-15 | 2015-04-22 | 富士通株式会社 | Method executed by video decoding apparatus |
| CN102833533B (en) * | 2011-06-15 | 2015-05-13 | 富士通株式会社 | Video decoding apparatus and video decoding method |
| CN106878733A (en) * | 2012-09-24 | 2017-06-20 | 株式会社Ntt都科摩 | Dynamic image prediction decoding device and dynamic image prediction decoding method |
| CN106878733B (en) * | 2012-09-24 | 2020-01-07 | 株式会社Ntt都科摩 | Moving picture predictive decoding device and moving picture predictive decoding method |
| CN104104961A (en) * | 2013-04-10 | 2014-10-15 | 华为技术有限公司 | Video encoding method, decoding method and apparatus |
| US9706220B2 (en) | 2013-04-10 | 2017-07-11 | Huawei Technologies Co., Ltd. | Video encoding method and decoding method and apparatuses |
| CN104104961B (en) * | 2013-04-10 | 2018-09-21 | 华为技术有限公司 | A kind of method for video coding, coding/decoding method and device |
| CN113596475A (en) * | 2021-06-24 | 2021-11-02 | 浙江大华技术股份有限公司 | Image/video encoding method, apparatus, system, and computer-readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| JP5394463B2 (en) | 2014-01-22 |
| JP2012044705A (en) | 2012-03-01 |
| JP2014007758A (en) | 2014-01-16 |
| JP5650294B2 (en) | 2015-01-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103024382B (en) | Dynamic image encoding device, dynamic image encoding method, dynamic image decoding device and dynamic image decoding method | |
| CN101218829A (en) | Dynamic image encoding device, dynamic image encoding method, dynamic image encoding program, dynamic image decoding device, dynamic image decoding method, and dynamic image decoding program | |
| KR102252816B1 (en) | Method and apparatus for determining depth of encoding unit | |
| CN102640496B (en) | Picture coding device, picture decoding apparatus, method for encoding images and picture decoding method | |
| CN103079070B (en) | Video encoder | |
| RU2559738C2 (en) | Method and device for encoding/decoding of motion vector | |
| JP7261275B2 (en) | Method and apparatus for encoding or decoding video data in FRUC mode with reduced memory access | |
| CN101682777B (en) | Image prediction encoding device, image prediction decoding device, image prediction encoding method, image prediction decoding method | |
| CN103647972B (en) | Moving picture decoding method and moving picture encoding method | |
| CN103931190A (en) | Image coding apparatus, image coding method, image decoding apparatus, image decoding method, and storage medium | |
| KR20230057481A (en) | Intra-prediction mode concept for block-wise picture coding | |
| KR20110114524A (en) | Device, method and program for image prediction encoding, device, method and program for image prediction decoding, and encoding/decoding system and method | |
| CN109429071A (en) | Picture coding device, picture decoding apparatus and image processing method | |
| JP2020523818A (en) | Method and apparatus for encoding or decoding video data in FRUC mode with reduced memory access | |
| CN114900691B (en) | Encoding method, encoder, and computer-readable storage medium | |
| JP2009272702A (en) | Motion vector searching apparatus and control method thereof | |
| TW202032995A (en) | Encoding and decoding a picture | |
| JP2006020217A (en) | Image coder | |
| JP4429996B2 (en) | Moving picture coding program, moving picture coding method, and moving picture coding apparatus | |
| JP6992825B2 (en) | Video coding device, video coding method, video decoding device, video decoding method, and video coding system | |
| WO2023044916A1 (en) | Intra prediction method, encoder, decoder, and encoding and decoding system | |
| JP4184249B2 (en) | Image encoding device | |
| WO2023044918A1 (en) | Intra prediction method, coder, decoder, and coding/decoding system | |
| WO2011002091A1 (en) | Moving image decoding method and moving image decoding apparatus |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20080709 |