Motion-blurred Image Restoration Algorithm Research And Achievement

In the process of collecting image, there is a relative motion between the camera and the object in the very short time when the shutter opens and closes, which can cause the image to be motion-blurred. Image restoration is to get the effective information hidden in the blurred image, in order to get better visual effect image. Generally, the motion-blurred image restoration method is to estimate two key parameters of motion-blurred angle and motion-blurred length, then get the motion-blurred point spread function(PSF), finally use a deconvolution algorithm to deal with it. In the both two aspects of the blur parameters estimate and deconvolution, new algorithms are proposed.In the aspect of estimating motion-blurred parameters, in view of this problem that using a Fourier spectrum of motion-blurred image and Radon transform to estimate blur angle is sometimes inaccurate, we propose a algorithm that is based on a new use of secondary Fourier spectrum combined with bit plane slicing algorithm and Radon transform for accurate estimation of blur angle. After getting blur angle, we rotate the same angle according to the clockwise direction to get the blurred image whose blur direction is horizontal direction, and then use differential autocorrelation method to estimate blur length. Using the motion-blurred angle and the motion-blurred length as parameters to constitute the PSF of motion blur. Through a lot of computer simulation experiments and comparing with several other existing algorithms, our motion blur parameter estimation method has better accuracy and robustness.In the aspect of image deconvolution, to overcome the problem that restored image of using standard Richardson-Lucy(RL) algorithm has ringing effect near the edges, we put forward a modified RL algorithm using reference maps, it adopts two stages to deblur. In the first deblurring stage, the first reference map is estimated from the blurred image and an intermediate deblurred result is obtained. And then the second reference map which has more accurate edge description, is updated according to both the blurred image and the deblurred result of the first stage, which is very helpful to suppress the ringing around edges, the final restored image is got with the help of the second reference map, and we use Gaussian image prior as the regularization to improve the standard RL algorithm.Finally, on the Matlab platform, we respectively conduct experiments of artificial motion-blurred image and the real motion-blurred image restoration, the experimental results show that our method can better restrain ringing effect while preserving image details.