Efficient Algorithm For Restoration Of Motion Blurred Images

One of the most popular issues of image processing is image restoration from amotion-blurred image. It shows the study of great practical significance. At present,many motion blur restoration algorithm has a great amount of computation and cannot guarantee the quality of results. This paper analyzes the motion blur imagedegradation model. This paper proposes an improved fuzzy image parameterestimation algorithm and an improved recovery algorithm by Wiener filter in responseto these problems. And this algorithm has less computation,and better recovery resultsby combination of these two methods. This algorithm has a high recoveryefficiency.Radon transform method is currently being widely used method using fuzzyparameter estimation image. But as the image scale increases, the calculation speedbecomes very slow. Therefore, this paper presents a fast and high-precisionparameter estimation method.We make the original spectrum to threshold operationand extract portion of the spectrum of a linear regular block detection.we cancalculating the average after it. Because the image after extracted is smaller, so wecan improving the speed of radon detection. And performing a threshold operation toincrease the detection accuracy.In order to get a better result of restoration from a motion-blurred image withnoises. We used the radon transform to estimate the blurred length and angle of theblurred image. We also used secondary Wiener filtering method to estimate the powerspectral density of the image to avoid the value taken by the human to affect therecovery results. The last we used the windowed Wiener filtering method to recoverimage. This method suppresses effectively the ringing effect caused by using thetraditional Wiener filtering method. The experiments show that the parameters which estimated by this method are more precise and improve the noise immunity andstability of image restoration, suppress effectively the ringing effect. We can get abetter restored image by this method.