Abstract :
The direction chain code has been widely used in image coding and recognition for its simplicity and low storage requirement. However, the researches on how to use chain code for shape retrieval are few. The traditional retrieval methods, chain code histogram (CCH) and Minimize Sum Statistical Direction Code (MSSDC) did not take the distribution feature into consideration. In our previous work, three novel descriptors, chain code distribution vector (CCDV), chain code relativity entropy (CCRE), chain code spatial distribution entropy (CCSDE) are discussed which take full advantage of the statistical feature and the distribution, spatial and relativity feature. In this paper, we compared the five algorithms on a large set of shape database. The experimental results denote the efficiency of each method.
Keywords :
image coding; image recognition; image retrieval; shape recognition; statistical distributions; visual databases; chain code distribution vector; chain code histogram; chain code relativity entropy; chain code spatial distribution entropy; contour-shape recognition; contour-shape retrieval; direction chain code; image coding; image recognition; minimize sum statistical direction code; shape database; Computational intelligence; Computer security; Entropy; Image recognition; Image retrieval; Image storage; Information retrieval; Information security; Shape; Spatial databases; CCDV; CCRE; CCSDE; chain code;
