A method for effectively eliminating video garbled characters or squares

Abstract

A method for effectively eliminating garbled codes or squares in video images requires both reducing or eliminating garbled codes or squares in video images and ensuring the quality of the images. A balance must be considered when selecting the encoding bits of dif_mv_v. The number of encoded bits of the absolute value dif_mv_v of the difference of the horizontal direction motion vector selected by this method according to the corresponding range order is: 1, 3, 5, 7, 17, 19, 23, 25, 27, 29, 31, 33 , 35, 37, 39. The garbled characters or squares of the video image under a single background can not only be effectively eliminated by the method, but also retain a very good video image effect.

Description

technical field [0001] The invention belongs to the technical field of computer and image analysis and processing, and in particular relates to a method for effectively eliminating video garbled codes or blocks. Background technique [0002] When performing H.264 encoding on TI's Da Vinci platform, in the case of a single video background, the encoded video stream will appear in a lattice shape, resulting in garbled or square phenomena in the video image, which will make the encoding Video image quality suffers. By analyzing the code stream, it is found that the sequence with a single background has a large and irregular value of the motion vector (MV) obtained by the motion search. [0003] The motion estimation cost consists of two parts: one is the cost MV_COST of the motion vector MV, and the other is the difference SAD between the current block and the best matching system. In the motion estimation cost, SAD plays a leading role. Analyze the code stream and test It is...

AI Technical Summary

Problems solved by technology

If the MV of block B is large, the mean MV will be inaccurate, and the mean MV will be used to determine the moving starting point of the reference block, resulting in inaccurate data of the moved reference block, which will adversely affect the search for the best matching block

In the same way, when the median prediction is performed, the prediction of the median point will also be inaccurate, so that the search by the median point cannot find the best matching block

The prediction process is a cumulative process. If the current macroblock MV is inaccurate, it will affect the subsequent macroblock MV inaccuracy.

Therefore, for a sequence with a single background, the MV obtained by motion search will appear relatively large and irregular, resulting in garbled or square video images