Satellite image super-resolution reconstruction method and device of combined convolutional network

Abstract

The invention discloses an image super-resolution reconstruction method and device for a combined convolutional neural network, and the method comprises the steps: importing a source image into an automatic learning network, and obtaining a feature point neuron value image of the source image; updating and enhancing the neuron value of each feature point; degrading the quality of the enhanced feature point neuron value graph; wherein the low-resolution feature point neuron value graph is imported into a convolution network, the convolution network comprises at least three layers of convolution units which are connected step by step, each convolution unit is sequentially composed of a deconvolution layer and a convolution layer, and the magnification times of the three layers of convolution units which are connected step by step are increased step by step; and the convolutional network outputs a super-resolution image with a preset multiple. According to the method, image features of a target image are classified and learned through an ELM network structure, feature detail information of the image is enhanced, and a low-resolution image with enhanced feature information and detail information is reconstructed through a Laplacian pyramid network algorithm to obtain a high-resolution image.

Description

technical field [0001] The invention relates to the technical field of image enhancement, in particular to an image super-resolution reconstruction method and device with a combined convolutional neural network. Background technique [0002] With the current development of various observation technologies, the market demand for high-resolution images is increasing. Image super-resolution reconstruction technology uses a set of low-quality, low-resolution images to generate a single high-quality, high-resolution image. The application field of image super-resolution reconstruction is extremely broad, and it has important application prospects in military, medical, public safety, satellite remote sensing images, computer vision, etc. Due to the limitations of some optical imaging equipment and hardware conditions, the resolution of the obtained image cannot meet the requirements, and the resolution of the image can only be improved through post-processing and adjustment. Ima...

AI Technical Summary

Problems solved by technology

However, the current problem is that the neural network structure is single and can only be enlarged by a specified multiple, and the feature information and detail information of the target image are still not accurate enough.

Images