Image encoding and decoding method and system based on multiple description network

Abstract

The present disclosure discloses an image encoding and decoding method and system based on a multi-description network. For encoding: receiving image information, normalizing the image information, and obtaining convolution features through four-layer convolution operations. The obtained convolution features are divided into two descriptions, which are transmitted through different channels respectively. After two layers of convolution operations and quantization operations, two final multi-description code streams are obtained. For decoding: receive a multi-description code stream, if only one multi-description code stream is received, the deconvolution operation is performed on the received multi-description code stream through two edge decoder networks respectively, and the edge reconstruction information is obtained. If two multi-description code streams are received, the central decoder network simultaneously deconvolutes all the received multi-description code streams to obtain central reconstruction information. The denormalization operation is performed to obtain the edge reconstructed image and the center reconstructed image, respectively. The coding efficiency can be improved on the premise of ensuring that the packet loss problem is solved, and a reconstructed image with higher accuracy can be obtained.

Description

technical field [0001] The present disclosure relates to the field of communication technology and the field of computer technology, in particular to an image encoding and decoding method and system based on a multiple description network. Background technique [0002] The statements in this section merely mention background art related to the present disclosure and do not necessarily constitute prior art. [0003] With the development of multimedia technology and computer networks, the high-quality transmission of images and video signals in the network has become more and more important. The multiple description coding method can prevent the problem of image and video quality degradation caused by packet loss in image communication of a noisy channel, thereby saving transmission costs. Multiple description coding assumes that there are multiple paths between source and sink, and that the probability of simultaneous errors on each path is very low. By generating multiple ...

AI Technical Summary

Problems solved by technology

[0005] Multiple description encoding was proposed at the Shannon Theoretical Research Conference in September 1979, when Gersho, Ozarow, Witsenhausen, Wolf, Wyner, and Ziv posed the following question: If a source is represented by two separate descriptions, when these descriptions are What are the limitations on the quality of the sources when separated or combined

[0006] In the existing technology, it is impossible to achieve high-quality image compression at low bit rates, to improve the coding efficiency under the premise of packet loss, and to obtain high-precision reconstructed images.

Images