An Efficient and Parallelable Image Coding Method Oriented to Sparse Coefficients

Abstract

The invention discloses an efficient and parallelable image coding method oriented to sparse coefficients. The method includes the following steps that (1) a coefficient organizational method based on inter-frequency prediction is utilized, the coefficients are converted with the block as the unit and in the sequence that block energy is progressively reduced, a dense region and a sparse region are divided, entropy coding is executed on the dense region, and an entropy coding method based on small probability coefficient representation in the step 2 is executed on the sparse region; (2) the conversion coefficients of the sparser region are coded by adopting the entropy coding method based on small probability coefficient representation, and then image coding is finished. The method is high in coding efficiency and supports parallel processing and asymptotic transmission, and code streams can be cut off.

Description

technical field [0001] The invention relates to a high-efficiency parallel image encoding method oriented to sparse coefficients, belonging to the technical field of image / video encoding and decoding, in particular to the encoding of transformation coefficients with sparse distribution characteristics. Background technique [0002] The objects of video coding and image coding are mainly natural signals. Natural signals have a strong correlation in the spatial domain. To improve the coding efficiency, the spatial domain signal must be processed first, and the correlation should be eliminated or weakened before coding. Transform coding is a process of removing correlation. The basic idea of ​​transform coding is to transform the image information described in the space domain into another orthogonal vector space (transform domain). After the alternating transformation, the correlation between the coefficients is basically eliminated, and the energy is mainly concentrated on ...

AI Technical Summary

Problems solved by technology

The current organization method for sparse coefficients has certain limitations when choosing a data structure. For example, the combination of zero-tree structure and bit-plane coding for wavelet transform coefficients makes it difficult to achieve parallel processing. The organization of bit-plane coding does not make full use of the overall statistical properties of the transform coefficients, and requires the subsequent use of context-based adaptive arithmetic coding with high complexity to achieve better coding efficiency

Commonly used entropy coding algorithms are usually not suitable for directly compressing transform coefficients (such as wavelet transform coefficients) with low continuity and high probability of zero value, and it is difficult to achieve high compression efficiency. Adaptive arithmetic coding based on context has higher The coding efficiency is high, but the computational complexity is also high

Images