Research On Fast Image Matching Method Based On The Improved LSH Algorithm

Digital image matching technique is an important branch in the field of imageprocessing applications, which is now widely used in many fields, such as medicine,biology, and information processing, etc.. Digital image detection and recognition canmake people’s lives and work more convenient. Therefore, the study of fast imagematching method has important theoretical significance and application value.The researches on the image matching algorithm are mostly based on recall andprecision, which may make us ignore the matching rate of the algorithm. However,the image matching rate plays a very important role in digital image matchingprocessing application. In order to improve the image matching rate, starting from thematching method based on the color feature and the exiting image matching methods,this paper proposes a new fast image matching method based on color feature.The main contents and algorithm are as follows. Firstly, for the purpose ofimproving the image matching rate and ensure the accuracy, this paper proposes arevised hierarchical color histogram matching method. The main idea of this methodis to use the combination of the global color histogram matching method and thehierarchical color histogram matching method to match the images. This kind ofmethod can not only avoid the special disadvantages of the global color histogrammatching method, and also can work out the problem of hierarchical color histogrammatching method caused by its changeless reversal. The shortcomings of this methodlies in the low processing rate of over-sized image dimension and mass data, sofurther researches are needed. Secondly, LSH algorithm is applied in the fast imagematching method based on the hierarchical color histogram matching method. Aimingat the over-sized image dimension and redundant image database, this paper putsforward a dimensionality reduction of LSH fast image matching method. Thedimensionality reduction strategy adopted in this paper is the image gray processingusing the weighted average method. This method lower the image dimension, shortenthe projection time and therefore to improve the image matching rate.In conclusion, aiming at the shortcomings of the traditional methods, this paperproposes new fast image matching method, and then compares the performing of therevised method through experiment. The experiment shows that the new method can improve the image matching rate on the basis of ensuring the accuracy and can matchthe target images precisely.