FPGA based Passive Auto Focus System using Adaptive Thresholding

We present an FPGA based passive auto focus system which uses an adaptive thresholding to obtain well-focused images. Because the adaptive thresholding algorithm can evaluate the difference between each pixel and its neighborhoods, it can be applied to decide the specific area of an image contains high-frequency components or not. If the system decides a captured image has low-frequency components, the auto focus lens driver drives the lens until it obtains well-focused image. The algorithm we use is reasonably straightforward and does not require high computational power for real-time operation. The proposed system was implemented using FPGA and was successfully evaluated with a standard RS-170 video camera capturing 640 times 480 gray-level images at 30 frames per second with a 12.5-75mmF1.8TV lens