Identification And Restoration Of Motion-blurred Image

During generation, transportation and recording, the image quality will be decreased due to the influences, such as the atmosphere turbulence, the non-linearity of the sensors, the phase difference of an optical system and the relative motion between camera and object and so on. Restoration of a blurred image is a very challenging and important job of digital image processing. It aims to recover image from a degradation state and to make the processed image as informative as possible to the original image.A method proposed here is employed to identify the decisive parameters, such as the motion blur direction and the blur extent. The direction of the motion can be dealt with as unchanged during a short exposing time. Uniform linear motion blur is a common kind of blur in motion, because the motion of a changed speed and a non-linearity can be regarded as a composed result of a uniform linear motion. It is assumed that the original image obeys isotropy Markov process with rank one. This paper proposes an efficient identification approach for motion-blurred direction from motion blurred image with a constructed 3x3 directional differentiation matrix and directional differentiation using bilinear interpolation. A proper choice of differentiation and the method of "first identify cursorily and then carry out the refined identification" can help to identify the direction of the motion, well then, the blurred image can be rotated to a horzontal axis and the image restoration can be made out easily in one dimension. A correlation function between the pixels in the blur extent of a blurred image is applied to calibrate the blur extent. With these two parameters (the motion-blurred direction and the blur extent), we can get the PSF (point spread function) and complete the restoration by filters.The blur direction identification method in this paper has a smaller calculational quantity and higher precision. And the blur extent identification method does work not only for uniform motion blurred image, but also for accelerated motion or vibration blurred image. It can work out with a high precision result even the motion blurred image is corrupted with noises. All the algorithms proposed in this paper are improved by contrast with basic algorithms. Further research to explore a faster and more efficient algorithm of digital image restoration will be a promising job.