Gradient contour example dictionary and weighted adaptive p norm-based single image super-resolution reconstruction method

Abstract

The invention discloses a gradient contour example dictionary and weighted adaptive p norm-based single image super-resolution reconstruction method, which mainly comprises the following steps: performing up sampling on an input low-resolution image by using an interpolation method; completing estimation of a high-resolution gradient image through a gradient contour example dictionary-based neighbor embedding super-resolution reconstruction method; constructing a shape-adaptive similar block group; constructing a weighted p norm restriction of a singular value of the shape-adaptive similar block group, and adaptively adjusting a p value restricted by a norm of each similar block group according to the significance of an image region; constructing a reconstruction cost function, and performing iterative optimization solving to obtain a finally output high-resolutio image. The gradient contour example dictionary and weighted adaptive p norm-based single image super-resolution reconstruction method disclosed by the invention is relatively good in subject perception and relatively high in objective evaluation value. Therefore the invention is an effective single image super-resolutionreconstruction method.

Description

technical field [0001] The present invention relates to image super-resolution reconstruction technology, in particular to a gradient profile example dictionary and weighted self-adaptive p The invention discloses a norm single image super-resolution reconstruction method, which belongs to the field of digital image processing. Background technique [0002] In real life, limited by factors such as imaging system hardware equipment, imaging conditions, and information transmission conditions, people often cannot obtain high-resolution observation images. Image super-resolution reconstruction technology can break through these constraints without increasing hardware costs, and use one or more low-resolution observation images with complementary information to reconstruct a high-resolution image, enabling people to Can better understand the image content and make further analysis and utilization. Compared with multi-image super-resolution technology, single-image super-reso...

AI Technical Summary

Problems solved by technology

Learning-based methods can recover image details well, but the performance is highly dependent on the similarity between the training set and the test set

Images