运动模糊图像的判定与恢复

Identification and restoration of motion-blurred image

为消除图像由于摄取时的相对运动所造成的模糊和失真,将未被降质的原始图像视为各向同性的一阶马尔可夫过程,利用双线性插值和3×3方向微分算子,采用"先粗后细"的方法,从降质图像中粗略确定运动模糊方向范围后,再高精度鉴别其具体数值.然后将降质图像旋转至水平轴,将二维问题简化为一维问题.这种优化后的模糊运动方向鉴别方法具有鉴别范围大、稳定性好的优点,既克服了Yitzhaky方法的不足又减少了计算量,提高了精度.再利用模糊图像模糊带内像素点之间的相关性估计模糊尺度,由运动模糊方向和尺度这两个参数得到点扩散函数,然后用滤波的方法得到恢复效果良好的复原图像.

To remove the distortion caused by relative moving during taking pictures, assuming that the original image obeys isotropic Markov process with rank one, an efficient identification approach is proposed for motion-blurred direction from motion-blurred image with a 3 × 3 directional differentiation matrix and the bilinear interpolation. A proper choice of differentiation and the method of ＂first identifying cursorily and then carrying out the refined identification＂ can help to identify the direction of the motion, and then the blurred image can be rotated to a horizontal axis and the image restoration can be made out easily in one dimension. This optimized identification of motion blur direction has high precision and stability compared with Yitzhaky＇s method. A correlation function between the pixels in the blur extent of a blurred image is applied to figure out the blur extent. With these two parameters （the motion blur direction and the blur extent）, the PSF （point spread function） can be obtained and the restoration by filters can be carried out with good results.