Semantic segmentation combined multi-description coding method and system

Abstract

The invention discloses a semantic segmentation combined multi-description coding method and system. The semantic segmentation combined multi-description coding method comprises the steps that an original image and a corresponding semantic segmentation label serve as input to be transmitted to a multi-description feature generation network; the multi-description feature generation network generates two similar but not completely same descriptions by using the image context features; the generated two descriptions respectively enter a semantic segmentation coding network to be coded, and in thecoder, the descriptions are firstly up-sampled into an up-sampled image consistent with an original image in size, and then the up-sampled image and semantic segmentation mapping pass through a generation network together to generate a residual image, and the residual image and pixels of the up-sampled image are added to obtain a coarse reconstructed image; and the generated coarse reconstructedimage is transmitted to the multi-description decoding network through an information channel, and the semantic segmentation label is encoded by the standard encoder and then is transmitted to the multi-description decoding network, and a final reconstructed image is obtained through decoding. The consequence that the image reconstruction quality is not ideal due to the packet loss problem in thetransmission process is avoided, and the coding efficiency is better improved.

Description

technical field [0001] The invention belongs to the field of image compression, in particular to a multi-description encoding method and system combined with semantic segmentation. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art. [0003] Multiple description coding is a common error control technology in the field of image and video communication, and its main function is to reduce errors as much as possible during data transmission. The main idea of ​​multiple description coding is: an input source is coded into multiple descriptions and transmitted on the network, and one or more of these descriptions are used to realize source reconstruction. Clearly, the reconstruction quality improves as more descriptions are received. In the 1970s, Bell Laboratories proposed the channel separation problem, that is, the odd-even sampling of the signal to fo...

AI Technical Summary

Problems solved by technology

[0004] Because multiple description coding effectively solves the packet loss problem of traditional source coding in unreliable network transmission and the problem of image reconstruction and video reconstruction quality degradation caused by network delay, it has become a hot research topic.

Images