DSP-based Video Image Feature Extraction

Along with the development of video image processing, feature extraction has beenwidely used. But for the minutiae feature extraction issue, because the amplitude of minutiaefeature is relatively small and mainly added on the direction items, so for the whole image,the information is small and is difficult to reconstruct the minutiae feature when displaying.In the condition above, this paper presented a fast feature extraction method combiningsimulative and digital technique. Through designing a simulative video processing circuit toextract the minutiae feature from the direction items and increase the amplitude of feature.Then based on the platform of high speed DSP image processing platform extracted theminutiae feature parameters swiftly. In the application process of algorithm design, improvedthe Otsu's local iterative segment algorithm, presented an image-segment algorithm based onchanging gray scale and entropy. Then this paper processed the region growing by using therectangle-growing method constructed on the difference of gray scale, meanwhileaccomplished the external and internal feature extraction by adopting thinning iterativealgorithm and gray level co-occurrence matrix. According to the test result, compared withtraditional method, by applying the method presented in this paper, the second order momentof initial image decreased and the entropy increased, which proved that this system can fulfillthe task of extracting minutiae feature of character region and external and internal parameterseffectively in lesser error range, otherwise the speed of processing improved obviouslycompared with traditional method.