Regional hierarchical predictive coding method for quantized block compressed sensing

Abstract

In predictive coding of quantized block compressed sensing, a large number of low-efficiency candidates can result in higher complexity. In order to reduce coding distortion of quantized block compressed sensing, the invention provides a predictive coding method based on a context sensing candidate set and regional hierarchical correlation. After all blocks are observed at the same sampling rate, each block is predicted and quantized according to a scanning sequence from inside to outside, and a current observation vector selects an inverse quantization vector having a minimum mean square error with the current observation vector from a context sensing candidate set as a prediction vector of the current observation vector; and all the blocks are divided into one of three regions according to hierarchical correlation, differentiated quality factors are set for different regions through a block coding model, and a key region is endowed with a quality factor larger than that of a non-key region. Compared with the prior art, the predictive coding method comprehensively utilizes the spatial correlation and hierarchical correlation between observation vectors, improves the matching degree and reference value of prediction vectors, and obtains good rate-distortion performance in prediction coding.

Description

technical field [0001] The invention relates to image signal acquisition and compression technology, in particular to an observation vector predictive coding method oriented to quantization block compression sensing. Background technique [0002] The measurement end of quantized block compressed sensing divides the target image into several non-overlapping blocks, uses the block-level observation matrix to sample each block independently, obtains the observation vector of each block, and then performs prediction and quantization coding. If the compression ratio of the entire image needs to be changed, the measurement end can adjust the sampling rate or quality factor. Although the block operation improves the real-time performance of image processing, it ignores the spatial redundancy between blocks. If the original pixels of the current block and the original pixels of adjacent blocks have a strong spatial correlation, then the correlation of the observation vector obtaine...

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is: how to perform optimized observation vector predictive coding in quantized block compressed sensing, and reduce the coding distortion of the target image under the code rate constraint

Images