Image compression method, computer equipment and computer storage medium

Abstract

The embodiment of the invention discloses an image compression method, computer equipment and a computer storage medium, a code rate control module at a coding end can map a weight factor into a code rate control vector, and multiplies the code rate control vector with a sparse feature map to obtain potential feature representation of a specified code rate, and the potential feature representation is used for compressing the specified code rate. The quantization unit quantizes floating points in the potential feature representation of the specified code rate into integers to obtain an integer type potential feature representation, and the lossless coding module performs entropy coding on the integer type potential feature representation to obtain a binary code stream, so that only one target neural network image compression model needs to be trained, and the image compression efficiency is improved. And the compression ratio, the code rate and the reconstruction quality of the image can be adjusted at will only by adjusting the weight factor, so that a plurality of image compression models do not need to be trained, a computer device does not need to deploy a plurality of image compression models, and the hardware storage overhead of user equipment is greatly reduced. The number of elements of the sparse feature map is greatly reduced, the calculation amount of a subsequent module can be reduced, and calculation resources are saved.

Description

technical field [0001] The embodiments of the present application relate to the field of image processing, and in particular, to an image compression method, a computer device, and a computer storage medium. Background technique [0002] Image compression is an important technology in the field of signal processing and computer vision. It aims to reduce the number of binary bits required for digital image transmission and storage, and to maintain the reconstruction quality of the transmitted image as much as possible. In recent years, many deep learning-based neural network image compression methods have achieved better performance than traditional image compression methods such as JPEG and BPG. However, there are still some problems that need to be solved in order to make the neural network image compression method can be deployed in practical application scenarios. [0003] The neural network image compression method uses a pre-trained neural network image compression mod...

AI Technical Summary

Problems solved by technology

However, the pre-trained neural network image compression model has a fixed compression rate of the image, and the bit rate and reconstruction quality of the output image of the same input image are also fixed. This situation cannot be adapted to image compression. The actual demand, because in the actual demand, the image receiving end has diversified requirements for the bit rate and reconstruction quality of the output image, and also requires the image compression rate to adapt to the continuous change of the network transmission bandwidth, so as to facilitate image transmission

Obviously, the fixed compression rate, the fixed code rate and reconstruction quality of the output image cannot meet the diverse requirements of compression rate, code rate and image reconstruction quality in actual needs.

[0004] In order to obtain images with different compression ratios, different bit rates and reconstruction quality, one way is to train multiple different neural network image compression models, and each neural network image compression model corresponds to a compression rate, bit rate and image reconstruction respectively. quality, but this method requires training a large number of neural network image compression models, which not only increases the hardware storage overhead of the user device, but also requires a lot of labor to train multiple models, so this method is not economical and practical

Images