Font Size: a A A

Research Of Orthodontics Image Retrieval Based On Multi-Feature

Posted on:2012-10-15Degree:MasterType:Thesis
Country:ChinaCandidate:Y Y WangFull Text:PDF
GTID:2178330338494845Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
The information is increasing faster and faster in the 21th century, computers and internet have flown into people's working, learning and living fields. They are full of our life. Image retrieval is a hot research field which carries out in this situation. Image processing, pattern identification and database technology are integrated with this technology. And it also promotes others knowledge's development.Image retrieval based on text is not up to people's increasing needs. Then image retrieval based on content was carried out by some academician. Users have only to choose example image and the system automatically finish feature extracting, identifying and matching, which improve the efficiency of retrieval and make things convenient for orthodontics auxiliary diagnosis.This paper firstly introduces the background and meanings of the study and the present condition of CBIR. Then relative technologies of CBIR are explained in detail. Secondly, the paper mainly introduces image retrieval based on color feature and image retrieval based on shape feature. Because good retrieval efficiency can be gotten only by using single bottom image feature, the paper combines color feature and shape feature to carry out image retrieval. HSV color space is used and quantized at first. And color feature is described by combining cumulative histograms and spatial information. The image is divided into several parts, and weighted differently to get spatial information. The images are divided image by combining Canny operator and Otsu threshold method to get seven rectangular-shape features. Multi-feature method is to weight color feature and shape feature to take similarity matching.Finally, the experiments are carried out by using orthodontics images and some images from million images base with the above methods. The results of the experiments show that image retrieval based on multi-feature has a better effect, especially for orthodontics images. The method can be used in orthodontics auxiliary diagnosis.
Keywords/Search Tags:Image retrieval, Cumulative histograms, Image segmentation, Moment invariant
PDF Full Text Request
Related items