基于方向微分的运动模糊方向鉴别的改进算法

Improved Method of Directional Derivation to Identify the Direction of Motion Blurred Image

对运动模糊方向鉴别受图像中目标形状影响的情况,基于目标不可能同时出现在图像四角的考虑,提出鉴别图像四角替代整幅图像的思想,完善了现有的基于方向微分的加权平均法,扩大了其应用范围。同时,分两次设定方向角的搜索步长,减小了计算量。试验表明,改进后的基于方向微分的加权平均法运行时间为现有方法的29.27%,鉴别结果的均方误差为现有方法的65.98%。

The identification of motion blur direction is usually influenced by the shape of object. As one object can＇ t occupy four corners of the whole image, the method of using four corners to identify object instead of using whole image is brought forward, the weighted average method based on direction derivation is perfected and its application range is enlarged. This method sets the step-search of direction angle twice to reduce the computation. The experimental result shows that by the improved method the running time is reduced to 29.7% and the mean square error is reduced to 68.55%.