Font Size: a A A

Hierarchical Image Retrieval Based On Text And Content Information

Posted on:2015-03-01Degree:MasterType:Thesis
Country:ChinaCandidate:Y LiFull Text:PDF
GTID:2348330485995865Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
With the rapid development of multimedia technology and computer network,multimedia image information is also growing at an amazing speed. How to obtain the desired information quickly and accurately from the massive image database has become a hot topic in recent years. This paper takes a research on searching from massive news image database, proposing a method of quantitative color based on Gauss weighted distance.This paper mainly takes research on the retrieval method based on image text and content information. Researching method and breakthroughs are as follows:This paper takes studies on the description method of image based on color histogram, aiming at the defects by using the traditional HSV color space quantization method and proposing Gauss weighted distance quantization method which is able to reduce the quantization error. This paper takes studies on the retrieval method based on texture features, which statistics the frequencies of different texture of sub block.This combined retrieval method of using combined color and texture feature is better than using only single color features. This paper takes studies on the KD-Tree of image feature which is a rapidly and accurately algorithm searching the nearest high-dimensional vectors of specific feature. In experiment, it takes averagely 100 to200 ms querying the set of 80,000 images database.This retrieval system based on text and content information has been verified by experiment in XinHua News image database, which shows that this method can improve the performance of image retrieval system.With the fast development of handheld devices, the demand for handheld devices is increasing rapidly. This paper not only introduces improving the speed and accuracy of image retrieval, but also discusses querying and receiving the image result directly in embedded equipment. Considering the image transmission speed may affect the users' experience, this paper takes research on increasing the transmission speed of embedded equipment with double network card by using RAW socket to achieve load balancing as well as increase the network bandwidth.
Keywords/Search Tags:Image Retrieval, Gauss Weighted Dist, HSV, KD-Tree, Image Transmission
PDF Full Text Request
Related items