Image Motion Blur Mechanism And Recovery Technology And Methods Of Research

The image restoration is a important topic of the image processing, its main purpose is to improves the picture's quality, which studies how to recover the primitive image from the obtained degenerated image .Image restoration is an important task of the image processing, there have a broadly market and applying in practice.This article mainly study the restoration of straight line motion-blurred image and rotational motion-blurred image,compared restoration effect and give the conclusion. The main content of this paper are given below.1. Elaborated three algorithms about the restoration of straight line motion-blurred image: inverse-filtering, constrained least square method,wiener filtering. Introduced each algorithm's principle and formula separately, doing restoration experiments aiming at every method on condition of noise and no noise, compared the MSE value of restored image. The result indicated that the inverse-filtering can recover the primitive image basically,but it is quite sensitive to the noise. The restoration quality of constrained least square method has some improvement compared with the inverse-filtering in the non-noise situation, but it is still unable to eliminate the effective noise. The Weina filtering may eliminate noise by chooses the appropriate parameter.2. Elaborated two algorithms about the restoration of images blurred by rotational motion: a restoration approach based on geometric coordinate transformations and a restoration approach based on deconvolution along blurring paths.Introduced each algorithm's principle separately, doing the simulation experiment aiming at every method on condition of different degrees, compared the experimental result and the MSE value . The result indicated that when the rotation degrees is very small, the difference between two algorithms is not too big, when the rotation degrees is not very big, the latter algorithm surpasses the preceding algorithm, when the rotation degrees is very big, the quality of restoration image are both not very good.3. Propose a improvement to one kind of straight line parameter estimation law using the polar coordinate transformation thought, and apply it in the rotational motion-blur angle estimate, the experiment has proven that the method is feasibility. The experimental result indicated that this algorithm can estimate the rotation degree very accurately regarding not very complex image, when the degrees rotation is not very big and the image is complex, this algorithm can also estimate the rotation degree precisely, but the estimated value error is quite big when the degrees rotation is very big.