An Image Retrieval Technique Based on Entropy and Salient Block Matching
Tianwei Qu · Jisuanji fangzhen · 2009
This paper puts forward a new image retrieval method which combines image information entropy with salient block matching using salient points.At first,each image in the database is preliminary classified by computing information entropy compared with a given threshold.Secondly,salient points in an image are detected by harris.Next,the features of each salient block,a rectangle area including 5 ×5 pixels,are described by three important colormoments.Finally,the similar distance between two images are calculated based on counting the number of matched salient blocks between two pictures.Experimental results show that this scheme reduces retrieval time greatly,because the salient block in the algorithm is more comprehensive and precise as describing the visual information of a picture and the similarity measurement method is easy and highly effective.