Abstract: At present,Content Based points Retrieval is a subset of the CBIR．It tries to inquiry the trademarks similar to the sample image,based on the comprehension of image contents．CBIR in trademark has wide prospect in the stage of similar or same trademarks examination,even in the field of trademarks inquiry．This paper does mainly research on the extracting of the points characters and counting the similarity in images．Firstly,the image shape is extracted by Canny algorithm,then,the moments are introduced to image retrieval．Experiments show that the proved method has effective performance.

Key words: Image retrieval,Edge directions histogram,Invariant moments,Contour sequence moments