Lossless restoration of local blurred image based on deep residual network
Qing He, Chengshuang Miao · 2021
In the process of digital image transmission, it is easy to be affected by the environmental conditions, which leads to the local blur problem of digital image. Therefore, a local blur image lossless restoration method based on deep residual network is proposed. The identity mapping is set in the deep residual training network layer to obtain the structure of the new residual unit. Combined with the weight matrix of local fuzzy image segmentation, the edge of local fuzzy image is segmented. The fuzzy parameters of local fuzzy image are calculated, and the fuzzy wavelet method is used to calculate the fuzzy degree and fuzzy peak value of local fuzzy image. Finally, the lossless restoration formula of local blurred image is obtained to realize the lossless restoration of local blurred image. The experimental results show that, compared with the traditional restoration method, the restoration efficiency of this method is significantly improved, the highest restoration efficiency is 93%, and the restored image has higher definition.