Sparse Constrained Adaptive NLM Super-resolution Reconstruction Method for Text Images

Abstract

The invention relates to a sparse-constraint-adaptive NLM (non-local mean) super-resolution reconstruction method aiming at character images. The sparse-constraint-adaptive NLM super-resolution reconstruction method is a premise of the character image detection and recognition, and has wide application prospect in the aspects of smart city, video sensing of Internet of Things, and the like. The adaptive sparse constraint NLM (non-local mean) super-resolution reconstruction method solves the problem of setting an h parameter according to the experience in the existing NLM super-resolution reconstruction algorithm by utilizing a statistical and structural sparsity self-adaptive calculation weight regulating factor h of the character image. The sparse-constraint-adaptive NLM (non-local mean) super-resolution reconstruction method utilizes L-1 norm distance to measure the similarity of image sub-blocks, and avoids the excess smoothness of the reconstruction. The sparse-constraint-adaptive NLM (non-local mean) super-resolution reconstruction method can obtain optimal values of a search window parameter p and a comparison window parameter q by experimental analysis, thereby obtaining higher reconstruction precision and efficiently lowering the time complexity of the algorithm. The sub-pixel motion can be estimated through a matching method, and the sampling factor of the super-resolution is enhanced by 1-4 times. The reconstruction results are deblurred by an adaptive total variation algorithm, and the edge texture information of the image can be efficiently maintained when the noise and parasitic ripples are suppressed.

Description

technical field [0001] The present invention relates to a sparse constraint adaptive NLM super-resolution reconstruction method for text images, especially an adaptive h-parameter NLM (Non-localmean) super-resolution reconstruction method for low-resolution text images. Perception, smart city and other aspects have broad application prospects. Background technique [0002] Improving image quality is an important topic in the field of image processing. As an important technology to improve image quality, image super-resolution has become a research hotspot. High-resolution text images are the premise of text image recognition and other applications, and have broad application prospects in text forensics, Internet of Things video perception, smart cities, etc. An important part of identification. At present, on the one hand, due to the constraints of imaging equipment and other factors, the resolution of surveillance video images is low. In addition, factors such as weather,...

AI Technical Summary

Problems solved by technology

[0004] The present invention mainly solves the technical problems of low spatial resolution and poor recognizability of sequence low-resolution image characters existing in the prior art; difficult selection of parameters of weight gain adjustment factor h, search window and comparison window, etc.; provides One solves the problem that the h parameter in the super-resolution reconstruction algorithm of the existing NLM method is set according to experience; in the calculation of the weight matrix, it proposes to use the L-1 norm distance to measure the similarity of the image sub-blocks, avoiding the reconstruction result It can not only achieve high reconstruction accuracy, but also effectively reduce the time complexity of the algorithm; it can achieve high-precision sub-pixel matching; and it can effectively maintain the uniformity of the edge texture information of the image while suppressing noise and parasitic ripples. A super-resolution reconstruction method based on NLM method for text images

Images