利用拉氏算子鉴别运动模糊方向

Motion blur direction identification in motion blur image by Laplacian

提出一种新的鉴别运动模糊图像的运动模糊方向的方法 ,它利用拉氏算子对运动模糊图像进行无方向性的二阶微分 ,并求微分图像的自相关 ,发现自相关图像中数值较大的点 (鉴别点 )能够有效标示出运动模糊方向。选取适当数目的候选鉴别点 ,并利用聚类方法剔除其中的奇异点 ,得到鉴别点 ;过零频尖峰 (自相关图像的中心点 )画一条直线 ,计算各个鉴别点到该直线的距离 ,求距离和 ;改变直线方向 ,当距离和最小时 ,直线的方向即为运动模糊方向。数据实验表明 ,这一新的运动模糊方向鉴别方法 ,具有抗噪声能力强、适用范围广、计算量小、鉴别精度高、稳定性好的优点。

After the second differentiation of a motion blur image by Laplacian, and the autocorrelation computation of the differentiated image, the points with large absolute value in the autocorrelation image can reveal the motion blur direction of the primitive image. We selected the suitable number of points for identification, kicked out the singular points among them by clustering algorithms, then drew the lines across the center point of autocorrelation at various angle. The direction of the line, which had the least sum of distances between every identification point and the line, was just the motion blur direction. Experiment results show that this new method has strong robustness, low computation, high identification precision and good stability.