A Denoising Autoencoder Prior Information Algorithm for 6D Embedding for Color Image Super-Resolution Reconstruction

Abstract

The present invention provides a 6-dimensional embedded denoising self-encoding prior information algorithm for color image super-resolution reconstruction, comprising the following steps, step A: using denoising automatic encoding (DAE) as a color image super-resolution reconstruction The prior information means, by copying 3 channels, establish a 6-dimensional embedding denoising self-encoder prior algorithm model; Step B: Train a denoising network with 6-dimensional variables as input, and then use the network-driven high-dimensional prior matrix to embed The prior information of the color image is super-resolution reconstruction; Step C: iterative restoration stage, the intermediate color image is mapped to a 6-dimensional image, processed by the network, and the average operator is used to convert it back to a three-channel image. The high-dimensional prior algorithm of the invention solves the problem that basic data elements fall into a local optimal solution, and effectively overcomes instability. The algorithm has good performance and good visual inspection performance.

Description

Technical field [0001] The present invention belongs to the field of color image processing, specifically a 6-dimensional embedded denoising self-encoded priori information algorithm for color image super-resolution reconstruction, mainly used in color image super-resolution reconstruction. Background technique [0002] Image processing is a comprehensive marginal discipline on image processing. From its research method, it is learned from many disciplines, physics, physics, biology, physiology, psychology, electronics, computer science, etc. Contact, although there is a focus on but complement each other. Alternatively, various disciplines have been supported by new theories, new tools, and new technologies such as artificial intelligence, neural network, genetic algorithm, fuzzy logic, and thus progressing continuous progress. From its research scope, it intersects multiple majors such as pattern identification, computer vision, computer graphics, and there is a difference in r...

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