Study On The Algorithms For Content-based Image Reverse Search Engine

Reverse image search engine is a new image search engine. It can express the user’ssearch intent more intuition istive and help user to achieve the result objectively. Use imageto search image has become the direction of image search engine. In this paper, severaltypical search algorithms are analyzed. Then the characteristics and applicable range of eachalgorithm are discussed. The algorithm is optimized to adapt the image which is complex andrich in content and improve the accuracy of search results.This paper mainly introduces the content-based image retrieval algorithm, including thecolor-based, stripe-based, shape-based and local invariant features based image retrievalalgorithm. The advantages and disadvantages of each algorithm in practical application areanalyzed. Then the accuracy and robustness of image retrieval algorithm has been improved.The main contents are as follows:（1）Improved the image retrieval algorithm which is based on color feature. Byusing the superiority of the HSI space in a color representation combined with themethod of block extraction the algorithm of image retrieval which is based on colorfeature is ameliorated. By using the advantages of block algorithm, including distributionin the color space, color histogram representation and the connection between the imageof each block, image retrieval is made more convenient and accurate;（2）The performance of image retrieval algorithm which is based on low-levelfeatures such as GLCM edge histogram, Hu invariant moments, corner detection, etc, iscompared to facilitate its application.（3）Weighted multi-feature image retrieval algorithms and multi-feature fusionbased algorithm are studied as a result of the composite of similarity obtained thougheach low-level feature based algorithm.The statistical analysis of experimental results shows that multi-feature based imageretrieval algorithm is faster and more accurate than low-level feature-based algorithm. It ismore suitable for the image content based reverse image search engine.