|
The content-based image retrieval (Content-based Image Retrieval, CBIR) is to use the image's color, shape , texture and other characteristics of the image query , tried on in the understanding of the basis of the image content , similar images retrieved and examples . CBIR directly from the image feature extraction indexing , feature extraction and indexing by a computer automatically implemented to avoid artificial subjectivity described , greatly reducing the workload . The shape as an important feature of the image , has a good environmental invariance , generally does not change with the color change of the image , the brightness of the environment changes , it has been widespread concern . In this paper, the characteristics of traditional Freeman chain code (Freeman Chain Code, FCC) , and the same moment feature extraction algorithm learning , research and design of the image feature extraction , feature matching and image retrieval algorithm . First, a brief introduction to the traditional Freeman chain code and the definition of Hu invariant moments feature extraction methods , analysis of the shortcomings of traditional Freeman chain code features for image retrieval , while Hu invariant moments image translation , scaling, good and rotation invariance . Second , proposed an improved Freeman chain code (Improved Freeman Chain Code, IFCC) feature extraction algorithm to give an unique and rotation invariant feature chain code image . The index on this basis , the use of biological sequence alignment similarity matching image chain code to calculate the match score , and then get the image retrieval results . Experimental results show that the robustness of the image , the algorithm has a good anti-rotation , at the same time to obtain better recall and precision , better able to reflect the shape of the image features, has better retrieval effect . Finally , taking into account the diversity of image features , in order to better use and reflects the shape of the image characteristics , image retrieval algorithm based on the improved Freeman chain code and image retrieval algorithm based on Hu invariant moments effective combination . Improved Freeman chain code to as rough shape characteristics retrieve , guarantee a certain precision , and thus take advantage of the moment invariant features fine retrieval , in order to achieve a better sort . The experimental results show that this retrieval method than the IFCC characteristics algorithm to obtain the better retrieval effect .
|