A writing method for macro block data of reference frame in video encoding-decoding process

Abstract

The method of writing in macroblock datan of reference frame during video coding / decoding is to perform continuous write operation based on the macroblock datan order. Specifically, while writing some macroblock, the datan are written into the memory space of reference frame in continuous address according to the row and line order. On one hand, this results in the direct write operation of macroblock in the order without need of calculating practical storing address to ensure the effectiveness of the available buffering memory addresses. On the other hand, when the macroblock is read as the reference frame of the next frame, the datan in the system buffering arean are effective macroblock data, resulting in saving in data reading time and no buffering miss. Compared with available technology, the present invention has raised reference frame read / write efficiency and speed.

Description

technical field [0001] The invention relates to a method in the process of video algorithm encoding and decoding, in particular to a method for writing macroblock data of a reference frame. technical background [0002] In the video coding algorithm based on motion estimation, the input frame is often divided into several macroblocks, and the encoding and decoding process is carried out in units of macroblocks. For example, when decoding, the reference frame and motion vector are used to obtain the predicted frame, and the image residual is added to obtain the restored data. Specifically, it is as follows: firstly, the motion vector and the image residual are separated from the bit stream data, and entropy encoding is performed respectively; then the motion vector and the macroblock corresponding to the reference frame are subjected to motion estimation (motion estimation) , ME), and finally add the image residual, and save the restored macroblock data as the reference fram...

AI Technical Summary

Problems solved by technology

[0003] Because the frame data is stored in the row and column sequence address of the frame when stored in the memory, and for each macroblock, its internal data address is discontinuous, so the restored macroblock data is stored as the next When the reference frame of the frame, that is, when writing macroblock data, the pre-read memory address is often invalid

Moreover, when performing decoding motion estimation for the next frame, the operating system reads the memory in units of macroblocks, and the system buffers with the sequential addresses stored, which will cause the address of the data read to be invalid.

Therefore, when the operating system processes the reference frame, the read and write of the macroblock will cause additional cache misses (cache miss)

Images