基于八面梯度均方差的跟踪系统自动调焦方法

为了实现目标跟踪系统的自动调焦,针对其实时性、准确性的要求,提出了基于八面梯度均方差的自动调焦方法.该方法以像素周围8个方向的梯度均方差作为清晰度评价函数,对消除噪声及光线波动的影响效果很好,能够准确地计算图像清晰度.同时,制定了强制连续调焦与精确调焦相结合的调焦策略.通过对几种评价函数的研究仿真,验证了八面梯度均方差函数在单峰性、灵敏性和稳定性等方面的优越之处；通过反复进行自动调焦实验,说明了调焦策略在克服评价函数局部峰值和精确聚焦方面的良好表现.该方法已应用于目标跟踪系统的自动调焦.

Auto-Focusing Method Based on Octahedral Gradient Variance for Tracking System

For the real-time and accuracy requirements of target tracking system, an auto-focusing method based on octahedral gradient variance was presented in this paper to actualize the automatic focus of target tracking system. In this method, variance of the gradients of each pixel in eight directions was regarded as focus measure function, with a good capability on eliminating the effect of noise and light fluctuations and accurately calculating the sharpness of image. Beside, a combination of mandatory continuously focusing and accurately focusing was chosen as focusing strategy. Through several studying and simulating, the mentioned function was validated better than others in following aspects, such as, single peak, sensitivity and stability and so on. Repeating auto-focusing experiments indicated the good performance of the focusing strategy to overcome local peaks and precise focusing. This method has been used in the automatic focus of target tracking system.