Image Matching Algorithm Research And FPGA Implementation

Video image matching is that mage information of different frames in a scene similar to the video image is compared, and with different degree of similarity between these images, a small part among image matching. It is widely used in object recognition, motion analysis, image stitching, video stabilization, etc., which is the basis of these image analysis techniques.For the images to be compared, we should first look for the features of the images which will have different way to describe; and next we should distinct these descriptions to get rid of the impact we judge the similarity of interference information, such as lighting, rotation. Therefore, on the basis of various studies on the image matching algorithm, we choose the image matching algorithm based on SIFT features extraction, and based on the original algorithm above, considering the hardware conditions (FPGA implementation), and without affecting the efficiency of the matching conditions, we improved the complexity of the algorithm, making a great breakthrough in real-time.In this paper, the image signal captured by the CCD was processed and computed, then was output and displayed ultimately with the matching point of the image signal to HDMI format in this system. FPGA implementation section mainly includes real-time video image capture module, a video format module, a storage module, SIFT algorithm for FPGA implementation module and feature matching module. Because of the simplified calculation algorithm SIFT still complex, taking into account the requirements of the hardware resources, the paper system is used Xilinx's Spartan-6 FPGA core processing platform type. At the same time, the core of the SIFT algorithm for hardware implementation module for the corresponding optimization algorithms to meet the requirements of real-time and high precision.Finally, in closing remarks, we summarize the entire project.