Image compression method based on discrete Gaussian mixture hyper-prior and Mask and medium

Abstract

The invention relates to an image compression method based on discrete Gaussian mixture hyper-prior and Mask and a medium, and the method comprises the following steps: carrying out the preprocessing of a to-be-compressed image, and obtaining a preprocessed image; extracting a feature map of the preprocessed image, generating a Mask value based on spatial feature information of the preprocessed image, and performing point product processing on the feature map and the Mask value to obtain hidden variable representation; adopting a plurality of Gaussian distributions to extract distribution conditions represented by hidden variables, and generating discrete Gaussian mixture hyper-priori values; quantizing the hidden variable representation, and performing entropy coding compression on the quantized hidden variable representation based on the hyper-priori value to obtain coding information of the compressed image; and decoding based on the coding information of the compressed image to obtain a reconstructed image. Compared with the prior art, the method has the advantages of good compression quality, high image compression efficiency and the like.

Description

technical field [0001] The invention relates to the field of image compression, in particular to an image compression method and medium based on discrete Gaussian mixture super prior and Mask. Background technique [0002] With the rapid development of the information age, the Internet produces a large amount of image information. Hundreds of millions of images are uploaded and downloaded by major Internet platforms every day. Therefore, how to store and transmit images more effectively to meet the needs of the times is a very challenging and very important task. [0003] Until today, several classic traditional image coding standards, such as JPEG, WebP, and HEVC, have been developed for decades. Most image service websites and application software use the JPEG compression algorithm. In recent years, due to the powerful feature extraction and expression capabilities of artificial neural networks, image compression algorithms based on deep learning have shown great potent...

AI Technical Summary

Problems solved by technology

However, the entropy model of the current deep learning-based image compression algorithm has limited ability to accurately fit the compressed representation, which limits the performance of the compression model.

In addition, there is still room for improvement in the ability of the existing deep learning-based encoders to extract features

Images