Super-resolution electronic device detection of clothing texture image based on improved dilated convolutional neural network

Convolutional neural networks cannot effectively reduce the one-to-many pathological problem in the super-resolution reconstruction task of a single image, and its role in improving the quality of image reconstruction is limited. We improved the feature extraction part by using resnet34 network with deeper network depth and less computation to replace vgg19 network, which further improved the feature extraction effect. Experimental results on mainstream datasets show that compared with the existing SSIR algorithm, the proposed algorithm greatly improves the image reconstruction accuracy and obtains the best peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) performance. The experimental results show that the reconstructed images have high PSNR and SSIM, and the parameters of the model are reduced to 1/29 of the depth residual channel focus network model.