A Fast Template Matching Based On Quaternion Zernike Moments

Quaternion moments as a very effective color image feature extraction technique, due to their advantages on the image representation, invariance and noise robustness, have been widely used in the field of image processing and achieved good results. The template matching technique is mainly used for target identification and tracking. Taking into account the complexity of the information contained in the color image, applying the quaternion moment and its invariants into the template matching and a new search matching strategy, we propose a new fast template matching algorithm that improves the matching speed in the premise of ensuring the accuracy of the matching to process color images. The main contents of this thesis are as follows:(1) This thesis introduces the background and the development of image processing, analyzes and summarizes the research process of the quaternion moment invariants and the template matching technology, and the research contents of each stage.(2) Taking the orthogonality of moments as cut-in point, the existing moment functions are introduced, the definition and simple properties of non-orthogonal and orthogonal moments are given, and through simple deduction the rotation invariant of some moments are given. In addition, the characteristics of different moments in practical applications and the comparison between them are briefly described.(3) We introduce the color image quaternion representation, and the construction process of the quaternion Zernike moment invariants is given. Through the experiments, the effect of the size of the base plate on the experimental results is studied and the generally better size of the backplane in the image normalization is obtained. At last, the effect of the moment invariants in color image processing is verified by experiments.(4) The existing template matching algorithms are introduced. We analyze and summarize the fast template matching algorithms from two aspects: the matching method and search strategy, and illustrate their advantages and disadvantages in application. For color image, we propose a new fast template matching algorithm, and the main idea is that we use quaternion Zernike moment invariants as the image feature through the search strategy from coarser region to finer region, and calculate the similarity between the searching subgraph and template image by calculating their relative distance. Finally, the experimental results show that the algorithm improves the matching speed in the premise of ensuring the accuracy of the matching, and the robustness to noisy is also testified.