Deep-convolution-neural-network-based super-resolution reconstruction method of single image

Abstract

The invention discloses a deep-convolution-neural-network-based super-resolution reconstruction method of a single image. The method comprises: step one, carrying out pretreatment; to be specific, converting an inputted image from an RGB color space to a YCbCr color space and only take a Y channel; step two, carrying out down sampling on the image after the pretreatment at the step one and carrying out interpolation by two channels to form training data of the channel 1 and the channel 2; step three, carrying out dense small block extraction on the image after the pretreatment at the step one and using a result as a tag; step four, with a gradient descent method, and a reverse conduction algorithm, optimizing a network model continuously by using the combined training data of the channel 1 and channel 2 as the input of a deep convolution neural network model and the tag as the output of the deep convolution neural network model; and step five, inputting a low-resolution image, carrying out dual-channel interpolation and then outputting a high-resolution image by using the trained deep depth convolution neural network. The method has advantages of light structure and good recovery quality.

Description

technical field [0001] The invention relates to an image super-resolution reconstruction technology, in particular to a single image super-resolution reconstruction method based on a deep convolutional neural network, which belongs to the field of digital image processing. Background technique [0002] Due to the limitation of camera hardware equipment or the influence of camera conditions, the image quality will be low and the edges will not be obvious. Improving the imaging accuracy of hardware equipment unilaterally will increase the cost of the product, and it cannot completely solve the interference of the imaging environment. The proposal of super-resolution reconstruction technology is undoubtedly the best way to solve this problem, which can not only reasonably avoid the waste caused by improving the imaging system, but also effectively improve the image quality. [0003] The method based on sparse coding is one of the representative methods for image super-resoluti...

AI Technical Summary

Problems solved by technology

The method based on sparse coding requires a lot of preprocessing of the input data, paying special attention to the learning and optimization of the dictionary, and each step needs to be optimized one by one. If all steps are not treated as a complete framework for unified optimization, it may face complex structure and calculation speed. slow challenge

Images