Motion-blurred Image Restoration Theoretical Analysis And Implementation

Welding technique plays a very important role in industry. It is now developingtowards automation and mechanization. With CCD camera as its eyes and guided bythe resulting video feed,arc welding robot can track and weld automatically.However, because of the influence of all kinds of noise during the process ofacquiring the welding images and the vibration of the image, there are noise andmotion blur in the image.Based on the above issues, this paper is to study the pre-filtering enhancementand motion blurred image restoration algorithms and solve the problem of processingthe resulting image feed from the CCD camera of arc welding robot.For the image contained noise, the first thing we must do is to filter noisethrough image pre-filtering. This paper mainly studied the Spatial filteringenhancement and Frequency domain filtering enhancement and had come to theconclusion that the median filter is the best way to preprocess the resulting imagefeed from the CCD camera of arc welding robot. There are two cases about therestoration of motion blurred image: If there is enough prior knowledge, we canmodel the original images, and to fit the blurred images according to the originalimages; If there is no prior knowledge,we can model the blurred images todescribe and find ways to reduce or remove the effect of the noise, this method iscalled estimation method.This paper adopts the former way, this is to say, we mustget the Point Spread Function.If we want to get the Point Spread Function of theblurred images, we need to estimate the fuzzy length and direction of images, that isalso called fuzzy parameters estimation. As for the estimation of the fuzzy length, thispaper has studied two methods, they are the estimation method based on frequencydomain and the transform method based on Radon; As for the estimation of the fuzzydirection, this paper has studied the transform method based on Hough and thetransform method based on Radon and obtain a improving method to the estimationof the fuzzy length and get the best way to the estimation of the fuzzy direction. Afterfuzzy parameters estimation, we can get the Point Spread Function according to model of the motion blur image and restore the blurred-image by adopting thecorresponding image restoration algorithm. This paper has studied four differentimage restoration algorithms, they respectively are: Inverse filtering, Wiener filtering,Lucy_Richardson and constrained least squares method. However Wiener filtering isthe best way to restore the images acquired by the Arc welding robot according tosimulations.In the course of this study on image preprocessing, parameter estimate of theblurred-image and restoration algorithms, I mainly adopt MATLAB software tosimulate.