A gradient threshold auto-focus algorithm

Traditional gradient auto-focus algorithms have large amount of calculation which will cause the reduction of real-time performance. These algorithms are also weak in anti-noise capability which will result in the decline of unimodality and sensitivity. So a gradient threshold auto-focus algorithm was proposed to improve the focusing performance to meet the requirements of real time and accuracy in auto-focusing subsystem of photoelectric tracking system. The proposed algorithm took the local variance as a local threshold to distinguish the edge pixels from non-edge pixels. Then it used a kind of new standard deviation of the whole image as a global threshold to weaken the effects of noise and background. At last, it used one of traditional gradient auto-focus algorithms to calculate the focusing value of the pre-processed image for clarity-evaluation. The results of lots of experiments show that the proposed algorithm has good real- time performance, strong unimodality, high sensitivity and powerful anti-noise capability. When the proposed algorithm is used in the auto-focusing subsystem of photoelectric tracking system, all the attractive performances remain, which traditional gradient auto-focus algorithm can't achieve.