Six-dimensional embedded de-noising self-coding prior information algorithm for color image super-resolution reconstruction

Abstract

The invention provides a six-dimensional embedded de-noising self-coding prior information algorithm for color image super-resolution reconstruction. The color image super-resolution reconstruction method comprises the following steps: step A, by using de-noising automatic coding (DAE) as a priori information means for color image super-resolution reconstruction, establishing a six-dimensional embedded de-noising automatic coding priori algorithm model by copying three channels; step B, training a denoising network with a six-dimensional variable as an input, and performing super-resolution reconstruction on the color image by using prior information embedded in a high-dimensional prior array driven by the network; and step C, an iteration recovery stage: mapping the intermediate color image into a six-dimensional image, processing the six-dimensional image by using a network, and converting the six-dimensional image into a three-channel image by using an average operator. According to the high-dimensional prior algorithm, the problem that basic data elements are caught in a local optimal solution is solved, and instability is effectively overcome. The algorithm has good performance and good visual inspection performance.

Description

technical field [0001] The invention belongs to the technical field of color image processing, and specifically relates to a 6-dimensional embedded denoising self-encoding prior information algorithm for color image super-resolution reconstruction, which is mainly used in the field of color image super-resolution reconstruction. Background technique [0002] Image processing is a comprehensive marginal subject about image processing. From the perspective of its research methods, it learns from and interacts with many disciplines such as mathematics, physics, biology, physiology, psychology, electronics, and computer science. Links, although each has its own emphasis, are complementary to each other. In addition, each of the above disciplines has been supported by new theories, new tools, and new technologies such as artificial intelligence, neural network, genetic algorithm, and fuzzy logic, and thus has been continuously developed. From the perspective of its research scop...

AI Technical Summary

Problems solved by technology

[0011] Existing image super-resolution methods often provide overly smooth reconstructed images, with loss of details and residual ringing

Images