Font Size: a A A

Image Retrieving System Research Based On Content Feature

Posted on:2007-05-02Degree:MasterType:Thesis
Country:ChinaCandidate:X M GouFull Text:PDF
GTID:2178360215981622Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
With the rapid development of multimedia and Internet technology since the 1990s, a hugeamount of digital images have been ever-increasingly produced. It is very important to manage theenormous size of image databases and retrieve images we needed. The drawbacks of traditional text-based image retrieval are time consuming and labor consumption. When a text-based image is indexed,this method can not represent contents of the image. Particularly, different persons may perceive thesame image differently. In specific application, this cannot meet the user's requirements. Thus theCBIR technology, which utilizes the content of image for similar query in image database, isincreasingly becoming an active research domain in recent years. The content of image involves color,shape and texture, which are the basic features and also the visual features in conformity with human.This thesis presents a practical image retrieving system based on color and texture characteristics.In CBIR system, feature extractions and similar measurement methods for the content of image areimportant. The color features, which we extract in this system, are a 72 bin vector of one dimensionby statistical color histogram in HSV space, a set of color average of sub-images, a set of position ofsub-images by dividing the original image and using statistic pixels color information, a set ofrepresentative color of an image using a suitable cluster algorithm in the HSV space. In order toextracting texture feature, we cut down the amount of calculation, compress the image and convert itto gray image. On this basis, the four numerical features identifying the content of image are elicitedfrom gray-level co-occurrence matrix of the image.Finally, we evaluate the retrieved result. Practicality of the system will prove the methodology isfeasible.
Keywords/Search Tags:Content-based image retrieval, feature extraction, color feature, texture feature, Similarity Matching
PDF Full Text Request
Related items