Image Quality Improvement of Motion-Blurred JPEG Image using Image Restoration Method

提出了一种改进运动模糊JPEG图像的方法。传统的正则化图像复原方法考虑了JPEG编解码过程模型、边缘方向和局部活动性。传统的正则化图像复原方法不考虑JPEG编解码过程，从时域改进的图像中提取边缘方向。计算机仿真结果表明了该方法的有效性。但运动模糊参数的估计是困难的，因此，代替估计的运动模糊参数，提出了评估的改进运动模糊JPEG图像。改进的运动模糊JPEG图像的评价值是通过对改进的运动模糊JPEG图像应用测量差分滤波器输出图像的频谱的偏斜度。改进的运动模糊JPEG图像的上述评估值基于测量微分滤波器像白化滤波器一样起作用的思想以及通过测量微分滤波器使适当改进的图像变成白色的思想。计算机仿真结果表明，改进运动模糊JPEG图像的评价值近似于原始图像和改进运动模糊JPEG图像之间的均方误差。

A method of improving motion-blurred JPEG image is proposed. The model about JPEG coding / decoding process, edge orientation and local activity are taken into account to conventional regularized image restoration method. Edge orientation is extracted from temporally improved image by conventional regularized image restoration method in which the JPEG coding / decoding process is not taken into account. The result of computer simulation shows effectiveness of this proposed method.Before improving process, the parameter of motion-blur is necessary. But estimating the parameter of motion-blur is difficult Therefore instead of estimating the parameter of motion-blur, evaluating the improved motion-blurred JPEG image is proposed. The evaluation value of the improved motion-blurred JPEG image is the skewness of the spectrum for the output image by applying measurement differentiating filter to the improved motion-blurred JPEG image. The above evaluation value of the improved motion-blurred JPEG image is based on both the idea that the measurement differentiating filter acts like the whitening filter, and the idea that suitably improved image is made to be white by the measurement differentiating filter. The result of computer simulation shows that the proposed evaluation value of the improved motion-blurred JPEG image acts like mean square error between the original image and improved motion-blurred JPEG image.