Abstract
Template Matching is a technique to identify the parts on an image or subsequences of a sequence to match a prede ned template. It can be used in digital image processing. There are many variations of the template matching problem. In this thesis, we have presented a matrix-based template matching algorithm. The algorithm utilizes a hash function to transform an image into a 2-D matrix. The hash function is rotation and re ection invariance. Then check the image blocks that have the same hash values to identify all the exact same templates between two images. And the method has implemented on GPU. The experimental results show that the performance on GPU is accelerated approximately 25 times than the CPU version.