CN1479308A - Storage access method - Google Patents
Storage access method Download PDFInfo
- Publication number
- CN1479308A CN1479308A CNA021422214A CN02142221A CN1479308A CN 1479308 A CN1479308 A CN 1479308A CN A021422214 A CNA021422214 A CN A021422214A CN 02142221 A CN02142221 A CN 02142221A CN 1479308 A CN1479308 A CN 1479308A
- Authority
- CN
- China
- Prior art keywords
- memory page
- page
- block
- read
- storage
- 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.)
- Granted
Links
Images
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The method includes following steps. Largest height and largest width of the prediction block is determined. Partial data in previous storage page adjacent to at least one borderline of at least storage of one page is stored repeatedly. Based on location of data block to be read, a data block on the storage page, where storing operation will not be occurred at borderline to change page repeatedly, will be selected to read. When a data block to be read, the invention eliminates operation of changing page by using repeated storing operation so as to increase bandwidth of storage and speed up reading speed.
Description
Technical field
The present invention is about access method of storage, the access method of storage of the video decoding system of the action of skipping when particularly reducing data read about the mode of utilizing the repeated storage data.
Background technology
In many modern video decode systems, as MPEG-I, MPEG-II and H261 etc., generally use compress technique to predict or the action of interframe encode, wherein motion compensation is based on block (Block), and each prediction block (Predicted Block) is all in conjunction with action vector (Motion Vector).The operation of motion compensation mainly is by on the reference picture, and the data vectorial according to action read prediction block.Generally speaking, the storer of prediction reference image is quite big, and is stored in the dynamic RAM (DRAM).Dynamic RAM is made up of many thesauruss, and thesaurus is made up of many storage lines, when before a storage line access data, must some extra time periods of experience, and as precharge and activate a storage line etc.If, will produce some extra time periods, the action of for example skipping so present accessed row is different from when previous.And this action of skipping can reduce the bandwidth (Bandwidth) that is read prediction block by dynamic RAM greatly.
In order to reduce the action of skipping, prediction block must be distributed in less memory page as far as possible, and prediction block must connect a page mode with one page and reads.Fig. 1 is a kind of synoptic diagram with block mode stored reference view data.As shown in Figure 1, reference picture is of a size of the 720*576 pixel, and each macro zone block (Macro-block) is the 16*16 pixel.If the memory page size of dynamic RAM is 1024 bit groups, then each memory page can be stored four macro zone blocks.The storage mode of Fig. 1 is each memory page (thicker solid line) storage 4 vertical macro zone blocks (thinner solid line).When reading the prediction block A of dotted line, data area A1, A2, A3, the A4 of this prediction block A is distributed in four different memory pages, and three actions of skipping so must be arranged.
Read prediction block though connect the order of one page with one page, can eliminate the action of skipping substantially, on bandwidth applications,, remain the big bottleneck on the bandwidth of dynamic RAM as the HDTV video decoding system.Especially on the frequency applications that the instantaneous decoding restriction is arranged, because in the motion compensation process, storer reads to need to reach fast in a large number, the action of too much skipping may cause reaching the demand of instantaneous decoding.
Summary of the invention
Because the problems referred to above, the purpose of this invention is to provide a kind of access method of storage, the data by repeated storage memory page border form the analog storage page or leaf, so can reduce or avoid the action of skipping fully.
For reaching above-mentioned purpose, access method of storage of the present invention, the action of skipping when reducing or eliminating data read with the partial data on repeated storage memory page border.This method comprises the following step: the maximum height and the breadth extreme that determine a prediction block; Storing step is at the partial data of the adjacent last memory page of at least one border of at least one memory page repeated storage; And read step, according to the position of the block desiring to read, select can on the border of repeated storage, not produce the memory page that skips and read this block.
Therefore, the present invention utilizes the mode of repeated storage to eliminate the action of skipping when reading block data, and then promotes bandwidth of memory, accelerates reading speed simultaneously.
Description of drawings
Fig. 1 is a kind of synoptic diagram with block mode stored reference view data.
Figure 2 shows that to merge the principle that two memory pages become an analog storage page or leaf, wherein (A) for block A for the location drawing, (C) that are positioned at memory page M page or leaf and N page or leaf in the location drawing of memory page M page or leaf, (B) for pixel of block A displacement to the right for being the memory page M_N page or leaf of simulating with the synoptic diagram of the boundary position data of the memory page M page or leaf of memory page N page or leaf repeated storage and (D).
Fig. 3 removes the example on all horizontal memory page borders for the present invention, and wherein (A) is that synoptic diagram, (B) of the overlapping storage of each memory page are the analog storage page or leaf.
Fig. 4 removes the example of all levels and vertical storage page boundary for the present invention, and wherein (A) is that synoptic diagram, (B) of the overlapping storage of each memory page is the analog storage page or leaf.
The drawing reference numeral explanation
The MB macro zone block
The A prediction block
M, N memory page
SP analog storage page or leaf
Embodiment
Describe the embodiment of access method of storage of the present invention in detail below with reference to accompanying drawing.
Figure 2 shows that to merge the principle that two memory pages become an analog storage page or leaf, wherein (A) for block A for the location drawing, (C) that are positioned at memory page M page or leaf and N page or leaf in the location drawing of memory page M page or leaf, (B) for pixel of block A displacement to the right for being the memory page M_N page or leaf of simulating with the synoptic diagram of the boundary position data of the memory page M page or leaf of memory page N page or leaf repeated storage and (D).
In Fig. 2 (A), memory page M page or leaf and memory page N page or leaf are the adjacent memory pages of two levels that is arranged in reference picture.Comprise because of prediction block A is stored a page M page or leaf fully, thus when system when reading prediction block A, do not have the action of skipping.Yet shown in Fig. 2 (B), when the prediction block A locations of pixels that moved right, the data line of prediction block A is stored in memory page N page or leaf.Therefore, when system when reading prediction block A, promptly produced from the M page or leaf and skipped to the action of skipping of N page or leaf.Therefore, generation for fear of the action of skipping, access method of storage of the present invention is at the partial data of the border of memory page N page or leaf repeated storage memory page M page or leaf, that is the data that the memory page N page or leaf data of being stored and memory page M page or leaf are stored overlap, as the hatched example areas of Fig. 2 (C).The peak width that overlaps is that the width of prediction block A subtracts 1.
Utilize this memory page overlapping technology, when system-computed goes out prediction block A and can stride across the vertical boundary of memory page M page or leaf, then can read the data of prediction block A, and comprise, so need not skip because of prediction block A is stored a page N page or leaf fully from memory page N page or leaf.So, during with the angle storer of system, can be the memory page M_N page or leaf of a simulation with memory page M page or leaf and N Pageview, shown in Fig. 2 (D).So as long as prediction block A is arranged in analog storage page or leaf M_N page or leaf, then system can not produce the action of skipping of vertical boundary when reading the data of this prediction block A.With the technology of this analog storage page or leaf, the size that can amplify memory page via the data of repeated storage between adjacent memory page.Certainly, Fig. 2 is for removing the example of vertical storage page boundary, the constructed horizontal memory page of the removal border that can be used for.
In addition, the size that depends on prediction block A about the area size of repeated storage.If the breadth extreme of prediction block is W
MaxWith maximum height be H
Max, unit is a pixel.For example, stored reference picture in MPEG I or MPEG II, the possible prediction block size of tool half-pix compensation comprises: 16 * 16,17 * 17,16 * 17 and 17 * 16.Therefore, breadth extreme and maximum height are all 17, that is W
Max=17 and H
Max=17.
Secondly, the capacity of supposing memory page is S
Page, and memory page storage macro zone block be shaped as rectangle, the pixel that defines its width and height is respectively W
PageAnd H
PageIf will remove horizontal memory page border the time, then shown in Fig. 2 (C), the line number H of the data that between the adjacent memory page of two levels, are repeated to store
DupBe (H
Max-1), so must satisfy H
PageH
MaxAnd if will remove the border of vertical storage page or leaf the time, the line number W that between two vertical adjacent memory pages, is repeated to store
DupBe (W
Max-1), so must satisfy W
PageW
MaxTherefore, can memory page be merged into bigger analog storage page or leaf, and reduce the action of skipping when reading prediction block according to memory size.
Fig. 3 removes the example on all horizontal memory page borders for the present invention, and wherein (A) is that synoptic diagram, (B) of the overlapping storage of each memory page are the analog storage page or leaf.In the specification of MPEG I or MPEG II, macro zone block is of a size of the 16*16 pixel, and the maximum height H of prediction block
MaxBe 17.If the capacity of memory page can be stored 4 macro zone blocks, then memory page height H
PageBe 64, so satisfy H
PageH
MaxCondition, can remove the horizontal memory page border that is positioned between vertical adjacent memory page.And the line number H that is repeated to store between each vertical adjacent memory page
DupBe 16, i.e. H
Max-1, just be the height of a macro zone block.So, with reference pictures store during to storer, except the memory page of first row does not need the repeated storage data, all the other each memory pages all repeat a memory page below the macro zone block data, as the hatched example areas of Fig. 3.Size with reference picture is that the 720*576 pixel is an example, and when not using storage means of the present invention, the line number of its memory page is 9 row.And after using storage means of the present invention, because the data that the memory page of every row must a macro zone block of repeated storage, so the line number of memory page increases by 1/3 approximately, that is total line number increases to 12.Though the capacity of storer needs to increase, but the elimination of level border shown in Fig. 3 (B), and is lifted at speed when reading prediction block.In Fig. 3 (B), the horizontal boundary of each memory page is eliminated, so the height of the defined memory page SP0～SP44 of system is extended.In other words, system is to increase the bandwidth that storage space exchanges storer for.Secondly, because of the memory page width W of Fig. 3 (A)
PageLess than prediction block breadth extreme W
MaxSo, can't remove the vertical storage page boundary.
Fig. 4 removes the example of all levels and vertical storage page boundary for the present invention, and wherein (A) is that synoptic diagram, (B) of the overlapping storage of each memory page is the analog storage page or leaf.In the specification of MPEG I or MPEG II, macro zone block is of a size of the 16*16 pixel, and the maximum height H of prediction block
MaxBe 17.If the capacity of memory page can be stored 16 macro zone blocks, then memory page height H
PageWith W
PageBe 64, so satisfy H
PageH
MaxWith W
PageW
MaxCondition, can remove the horizontal memory page border that is positioned between vertical adjacent memory page and remove the vertical storage page boundary that is positioned between the adjacent memory page of level.And the line number H that is repeated to store between the adjacent memory page of each vertical and horizontal
DupBe 16, i.e. H
Max-1, just be the height and the width of a macro zone block.So, with reference pictures store during to storer, except first first memory page that is listed as of going does not need the repeated storage data, memory pages of all the other each row all repeat to store a memory page below the macro zone block data, and memory page the most right-hand macro zone block data of all repeating to store last memory page of each row, as the hatched example areas of Fig. 4.Size with reference picture is that the 720*576 pixel is an example, when not using storage means of the present invention, the line number of its memory page be 9 the row and columns be 11.25 row.And after using storage means of the present invention, because the data that the memory page of every row and every row must a macro zone block of repeated storage, so the line number of memory page increases by 1/3 approximately with columns, that is always line number increases to 12, and always columns increases to 15.Though the capacity of storer needs to increase the border that can eliminate each memory page.Shown in Fig. 4 (B), the defined memory page SP0 of system has only one, so system all do not have the action of skipping when reading any prediction block, and is lifted at the speed when reading prediction block.So, with method of the present invention with reference pictures store to storer, system will not have the action of skipping when reading any prediction block, so use in that high bandwidth is instant, can overcome and move the bottleneck problem that bandwidth faced because of skipping.
Though more than with embodiment the present invention is described, therefore do not limit scope of the present invention, only otherwise break away from design of the present invention, those skilled in the art can carry out various variations or change.
Claims (12)
1. access method of storage, with the block be unit from the storage access reference image data, comprise the following step:
Determine the maximum height and the breadth extreme of a prediction block;
Storing step is according to described maximum height and breadth extreme, in the segment boundary data of the adjacent memory page of at least one border repeated storage of at least one memory page; And
Read step, if the position of the prediction block desiring to read is positioned on the memory page border of described repeated storage data, then selection can not produce the memory page that skips and read this block;
Eliminate the action of skipping when reading block data by the mode of utilizing repeated storage, and then promote reading speed.
2. access to store method as claimed in claim 1, wherein said storing step is at the partial data of the adjacent last memory page of horizontal boundary repeated storage.
3. access to store method as claimed in claim 2, the height of wherein said memory page must be more than or equal to the maximum height of described prediction block.
4. access to store method as claimed in claim 3, in the wherein said read step, when the block desiring to read can stride across the horizontal memory page border of a memory page time, then read the data of this block from another memory page.
5. access to store method as claimed in claim 1, wherein said storing step is at the partial data of the adjacent last memory page of vertical boundary repeated storage.
6. access to store method as claimed in claim 5, the width of wherein said memory page must be more than or equal to the breadth extreme of described prediction block.
7. access to store method as claimed in claim 6, in the wherein said read step, when the block desiring to read can stride across the vertical storage page boundary of a memory page time, then read the data of this block from another memory page.
8. access to store method as claimed in claim 1, wherein said read step be at the partial data of the adjacent last memory page of vertical boundary repeated storage, and at the partial data of the adjacent last memory page of horizontal boundary repeated storage.
9. access to store method as claimed in claim 8, the height and the width of wherein said memory page must be respectively more than or equal to the maximum height and the breadth extreme of described prediction block.
10. access to store method as claimed in claim 9, in the wherein said read step, when the block desiring to read can stride across the vertical storage page boundary of a memory page time, then read the data of this block from another memory page.
11. access to store method as claimed in claim 10, in the wherein said read step, when the block desiring to read can stride across the horizontal memory page border of a memory page time, then read the data of this block from another memory page.
12. a memory storage methods, with the block be unit with reference pictures store to storer, comprise the following step:
The length and width Pixel Dimensions H and the W of decision prediction block; And
Partial data at the adjacent last memory page of at least one border of at least one memory page repeated storage.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB021422214A CN100403276C (en) | 2002-08-26 | 2002-08-26 | Storage access method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB021422214A CN100403276C (en) | 2002-08-26 | 2002-08-26 | Storage access method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1479308A true CN1479308A (en) | 2004-03-03 |
| CN100403276C CN100403276C (en) | 2008-07-16 |
Family
ID=34147974
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB021422214A Expired - Lifetime CN100403276C (en) | 2002-08-26 | 2002-08-26 | Storage access method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100403276C (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100446572C (en) * | 2004-12-28 | 2008-12-24 | 联发科技股份有限公司 | Method of decoding digital video and digital video decoder system thereof |
| CN101556554B (en) * | 2008-04-07 | 2011-04-13 | 联发科技股份有限公司 | Storage system and method thereof |
| CN102033797A (en) * | 2009-09-28 | 2011-04-27 | 佳能株式会社 | Information processing apparatus, method for controlling information processing apparatus |
| CN101436163B (en) * | 2007-11-16 | 2011-12-28 | 慧荣科技股份有限公司 | Non-volatility memory storage device and access method thereof |
| CN101520689B (en) * | 2009-04-17 | 2012-10-03 | 成都市华为赛门铁克科技有限公司 | Computer memory device control method, computer memory device controller and memory device |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4346377A (en) * | 1980-05-30 | 1982-08-24 | Eltra Corporation | Method and apparatus for encoding and generating characters in a display |
| US6138188A (en) * | 1998-09-28 | 2000-10-24 | Mentor Arc Inc. | Buffer management device and method for improving buffer usage and access performance in data processing system |
| FR2787669B1 (en) * | 1998-12-22 | 2001-03-02 | Thomson Multimedia Sa | ADDRESSING METHOD FOR STORING IMAGE BLOCKS |
-
2002
- 2002-08-26 CN CNB021422214A patent/CN100403276C/en not_active Expired - Lifetime
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100446572C (en) * | 2004-12-28 | 2008-12-24 | 联发科技股份有限公司 | Method of decoding digital video and digital video decoder system thereof |
| CN101436163B (en) * | 2007-11-16 | 2011-12-28 | 慧荣科技股份有限公司 | Non-volatility memory storage device and access method thereof |
| CN101556554B (en) * | 2008-04-07 | 2011-04-13 | 联发科技股份有限公司 | Storage system and method thereof |
| CN101520689B (en) * | 2009-04-17 | 2012-10-03 | 成都市华为赛门铁克科技有限公司 | Computer memory device control method, computer memory device controller and memory device |
| CN102033797A (en) * | 2009-09-28 | 2011-04-27 | 佳能株式会社 | Information processing apparatus, method for controlling information processing apparatus |
| US8711417B2 (en) | 2009-09-28 | 2014-04-29 | Canon Kabushiki Kaisha | Information processing apparatus, method for controlling information processing apparatus, and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100403276C (en) | 2008-07-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100527099C (en) | Enhancing performance of a memory unit of a data processing device by separating reading and fetching functionalities | |
| CN1154366C (en) | Frame memory and method for recording frame image signal in SDRAM | |
| JP4910576B2 (en) | Moving image processing device | |
| KR960009748A (en) | Method and apparatus for efficient addressing of DRAM with video enhancement processor | |
| EP1998569A1 (en) | Method for mapping image addresses in memory | |
| DE112006002241T5 (en) | Dynamically configuring the cache of a video encoder for motion compensation | |
| JPH08123953A (en) | Picture processor | |
| CN100356780C (en) | Image storing method for compressing video frequency signal decode | |
| KR100612414B1 (en) | Image data processing system and method for reading and writing image data | |
| CN100444636C (en) | Method for improving SDRAM bus efficiency in video decoder | |
| US20080044107A1 (en) | Storage device for storing image data and method of storing image data | |
| CN1479308A (en) | Storage access method | |
| CN101216933B (en) | An image data memory projection method, assess method and device | |
| US7777751B2 (en) | Tiled memory array for full search motion estimation | |
| US6753871B2 (en) | Memory access method | |
| JP3120010B2 (en) | Image decoding method and image decoding device | |
| US6178289B1 (en) | Video data shuffling method and apparatus | |
| US20040061704A1 (en) | Memory access method for video decoding | |
| MY133444A (en) | Motion compensated digital video decoding with buffered picture storage memory map | |
| CN1154048C (en) | Address processing method | |
| US7420567B2 (en) | Memory access method for video decoding | |
| CN1309257C (en) | Video decoded memory accessing method | |
| JP3119994B2 (en) | Image data processing method, storage device used therefor, and image data processing device | |
| CN1168318C (en) | Video information decompressing system with high access power to efficient memory | |
| JP5261694B2 (en) | Image processing memory |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CX01 | Expiry of patent term |
Granted publication date: 20080716 |
|
| CX01 | Expiry of patent term |