Efficient channel attention algorithm for super-resolution reconstruction of remote sensing image

Remote sensing imaging equipment generally faces the problems of long distance and low imaging resolution, which directly affects the quality and application effect of remote sensing images. In order to solve this problem, a super-resolution reconstruction network based on feature enhancement of efficient channel attention is proposed. It introduces image super-resolution reconstruction technology into the field of remote sensing image processing, uses the grouped convolution feature enhancement module to extract and enhance the features of the image, and then uses the attention module composed of efficient channel attention and asymmetric convolution in parallel to establish the relationship between different regions of the image, and reconstructs low-resolution remote sensing images to obtain idealized high-resolution images. Finally, the experimental results show that the peak signal-to-noise ratio and structural similarity of this algorithm on the WHU-RS19 test set are 28.70 dB and 0.7539, which are 0.19 dB and 0.0066 higher than those of the suboptimal method, respectively, and the details of the reconstructed image are more abundant.