Natural image super-resolution method based on expectation maximization algorithm

Abstract

The invention discloses a natural image super-resolution method based on an expectation maximization algorithm, which comprises the steps of (1) inputting a low-resolution mage; (2) carrying out interpolation on the image; (3) acquiring a hidden image; (4) cutting the hidden image into hidden image blocks; (5) acquiring a similar matrix of each hidden image block; (6) acquiring a dictionary of each hidden image block; (7) acquiring the mean and the covariance of each estimation image block; (8) acquiring the maximum posterior estimation value of each estimation image block; (9) acquiring a high-resolution image; (10) calculating a relative error; (11) judging whether the relative error meets a termination condition or not; (12) updating data; and (13) outputting an optimal high-resolution image. According to the invention, the expectation maximization algorithm is introduced into the field of natural image super-resolution, and abundant information for recovering image details is acquired, thereby being suitable for image super-resolution under complex conditions.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a natural image super-resolution method based on a maximum expectation algorithm in the technical field of natural image processing. The present invention super-resolves a low-resolution natural image to obtain a clear high-resolution natural image, so as to provide more accurate and comprehensive information for subsequent interpretation, target recognition and target detection of images and images. Background technique [0002] Image super-resolution technology refers to the process of reconstructing a clear high-resolution image from a single or multiple low-resolution images. Low-resolution images have lower spatial resolution, which affects a more comprehensive and clear description of the scene. The purpose of image super-resolution is to obtain high-resolution images, enhance and enrich the details of the scene, so as to provide more accurate and comprehens...

AI Technical Summary

Problems solved by technology

However, the disadvantage of this method is that in the construction of dictionary pairs, this method needs to collect a large amount of external training data, which is unrealistic. At the same time, there is an error in the representation between high and low resolution images, so that The overall effect of recovery is not very good

However, the shortcomings of this method are that the method only uses the local information of the image itself as the training data, and at the same time, the method does not fully exploit the prior information of the local image block, so that the information provided by the Gaussian process is insufficient. In some cases, the reconstructed image results are unstable, the details and edges are not restored very well, and the reconstruction quality of local areas is reduced.

Images