Sudoku based fine motion estimation data reading method
A motion estimation and data reading technology, applied in the fields of digital video signal modification, electrical components, image communication, etc., can solve the problems of low read and write efficiency, DDR2SDRAM memory consumes a lot of time, large data volume and irregular inter-line operation, etc.
Patent Information
- Authority / Receiving Office
- CN · China
- Current Assignee / Owner
- Publication Date
- 2016-04-20
Smart Images

Figure 1 
Figure 2 
Figure 3
Abstract
Description
technical field
[0001] The invention belongs to integrated circuit design technology and relates to a method for reading fine motion estimation data of Jiugongge. Background technique
[0002] When the H.264 high-definition video encoding core performs fine motion estimation operations, the amount of data in the search box is large and irregular and there are many cross-row operations, so the read and write efficiency is very low, and it will take a lot of time to write data into the off-chip DDR2 SDRAM memory. In order to meet the requirements of real-time encoding for high-speed access to large amounts of data, this patent proposes a macroblock reading method accessed in a fixed order, a method of on-chip caching of data in the form of a nine-square grid, and the use of address pre-judgment and early termination of redundant data, etc. method to increase the speed of fine motor access. Contents of the invention
[0003] The purpose of the present invention is to provide...
Examples
Embodiment Construction
[0025] The present invention is based on the Jiugongge fine motion estimation data reading method, comprises the following steps:
[0026] Step 1, the method of reading macroblocks accessed in a fixed order:
[0027] Due to the continuity of the image in the frame, the image macroblocks (each macroblock has 16 rows and 2 columns of pixel data) accessed in a fixed order during encoding can be used for address prediction and hit judgment. In order to improve the reading and writing efficiency of off-chip DDR2SDRAM, a read macroblock request can continuously read the data of 4 macroblocks from off-chip, and only need to latch the address once every 4 requests. Since the addresses are continuous, the base address range of each macroblock can be calculated in advance;
[0028] Two sets of on-chip DPRAM are used to cache brightness and chrominance data. The brightness data is stored continuously in the off-chip DDR2SDRAM according to the start address of a frame of images, and the ...