A single-view image deblurring method with 1D deconvolution along a motion-blurred path

Abstract

The invention discloses a single-viewpoint image deblurring method that performs one-dimensional deconvolution along the motion blur path, which includes the following steps: S1. Input the blurred image, and estimate the blur kernel according to the outline of the target in the blurred image; S2. The kernel is refined to obtain a single-pixel width blur path, and the length and angle of the blur path are calculated; S3. Along the blur path, Chebyshev polynomial interpolation is used to extract the converted image of motion blur along the horizontal direction; S4. For horizontal direction The motion blur image is expanded line by line to the width of a linear combination of powers of 2 to obtain the expanded motion blur image; S5. Perform a one-dimensional deconvolution operation on the expanded motion blur image line by line to obtain The deblurred converted image; S6. According to the blur path, inversely transform the deblurred converted image to obtain a deblurred inverted image that is consistent with the original blurred image.

Description

technical field [0001] The invention relates to the field of image restoration, in particular to a single-viewpoint image fast deblurring method and system for performing one-dimensional deconvolution along a motion blur path. Background technique [0002] In recent years, with the development of my country's industrial automation and machine vision, especially in the automated production line, in order to improve production efficiency, the product must be continuously detected and analyzed without leaving the assembly line. In a short exposure time, the shooting system The relative movement between the target and the target often leads to blurred images, which seriously affects the image quality, making subsequent detection and recognition extremely difficult. Therefore, it is urgent to study the method of real-time deblurring of moving target images, promote the development of target detection technology in the field of intelligent production, and promote the wide applicati...

AI Technical Summary

Problems solved by technology

[0003] The currently proposed fast deblurring algorithm, whether it is to transform the multivariate inverse problem into a simple one-variable problem, or to use mixed Laplace prior knowledge and alternate minimization, is based on the algorithm development of the two-dimensional image blur model. It takes several minutes and cannot meet the real-time requirements

In addition, there are ways to improve processing time by using high-performance GPUs or building image processing systems with multiple DSPs

Although these methods can solve real-time problems, they are expensive

Images