Research On Shape Based Algorithm For Image Retrieval

Content-based image retrieval (CBIR) means that making use of color, shape, texture and etc to query images, try to retrieval similar images with the example based on comprehending the content of image. CBIR extracts features and constitutes indexes from images directly, feature extraction and index constitution can be realized by computer automatically, which avoids the subjectivity of manual description, and the workload is decreased greatly. As a important characteristic of an image, shape has good environmental invariance, and generally not changes with the change of color and brightness. So it becomes to be the hotspot of people's attention gradually.This paper researched and designed feature extraction, feature matching, and retrieval algorithms of images through studying the traditional Freeman chain code (FCC) and the Hu moment invariants.First of all, the definitions of FCC and Hu moment invariants were shortly introduced. Then, expanded on the feature extractions of FCC and Hu moment invariants, analyzing the shortcomings of FCC used in image retrieval, and also found that the Hu invariant moments have good invariance in the image translation, scale and rotation.Secondly, an image retrieval algorithm based on improved Freeman chain code(IFCC) was proposed. Acquire an image chain code with unique and rotation invariance. Indexing based on the IFCC, we use the biological -sequences comparison based on Smith-Waterman algorithm to measure the similarity of the shape encoded with IFCC, calculate the score of matching, and then get the retrieval results. Experimental results showed this shape-based image retrieval algorithm had good robustness on rotated images, obtained good recall and precision, could incarnate the shape feature well, and had good retrieval effect.At last, considering the variety of image features, in order to incarnate the shape feature of an image, the algorithm based on IFCC and the algorithm based on Hu moment invariants were combined effectively. Coarse retrieval was carried through shape feature, which could ensure certain precision, and then did the retrieval through Hu moment invariants characteristic, which could achieve better sorting effect. Experiments showed that this retrieval method achieved better effect than the IFCC method above.