Restoration Of Motion Blurred Images

With the development of image acquisition and processing, images are being used everywhere. Images deliver information for people. Generally, there are two reasons resulting in blurred images. They are defocusing blurred images and motion blurred images. Defocusing blurred images are generated when people falsely focus the focal point on other sights rather than the target. Farther the distance between the focal point with the target, more blurred the image will be. The motion blurred images will appear on certain situation when relative movement happen between the camera and the target while people are pressing the shutter button.Motion blurred images are divided into two categories according to their causes. The first one is the camera shock, and the second one is the high speed of the running target. Discard the rotation and distortion in the exposure time; we take the translation in consideration. During the very short time of exposure, the translation can be considered as uniform and whichever pixel moves during the exposure time, it will be blurred in the image. Motion blurred images caused by camera shock and fast running target have different characteristics. For the first one, all pixels in the image are blurred, and in the later ones, only the running target gets blurred while the background pixels remain unchanged.In classical de-blurred methods, these pixels on the same image will be equally treated in traditional method. One of the traditional processes to eliminate the blur effects is using Wiener filer. However, owing to the different features between the blurred target and the still background, it is hard to get well restored images.A novel restoration method of motion blurred images is presented here. This method is aiming at the restoration of motion blurred images with blurred target and the standstill background, and trying to restore a clearer image than using classical Wiener Filter. Therefore, two optimization solutions are proposed here:(1) Making use of the profile of the target, and segment the original image into two parts:the one is the motion blurred part, and the other one is standstill background. The background will be abandoned, while the motion blurred part will be restored. The contour profile of the object should be confirmed firstly.(2)Based on the principle of motion blurring, the information of the background and boundary pixels is used to compensate for the boundary of the motion blurred image.The proposed algorithm is validated by experiments. In this paper, we will illustrate the method by dealing the same motion blurred picture with different methods. Then we will get different restored images. When judging the de-blurred images from the same motion blurred image, we can easily identify that the one using the proposed method is better. In this paper, the restored image will be estimated both subjectively and objectively using Mean Absolute Error and Mean Square Error.