Font Size: a A A

Design And Implementation Of Dynamic Image Filter

Posted on:2007-06-29Degree:MasterType:Thesis
Country:ChinaCandidate:Y R QiFull Text:PDF
GTID:2178360212475771Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the enhancement of the computer's data dealing ability and the progress of the multimedia compression and coding technology, various resources on the Internet is becoming abundant. People share the information it provides and feel convenient, but at the same time, it brings eroticism, violence and counteraction to us. People, especially teenages are contaminated by the "information rubbish" . In order to give us a cleaner cyberspace, Nowadays, many countries and organization are purifying the web.These days, the software to prevent the erotic images is either based on blockading Web addresses or filtering erotic texts. They act poorly at veracity and real time. In order to prevent the spread of eroticism effectively, the approach is carrying on research on content-based image filtering. Many experts have done much in this field and gain some achievement.In this paper, we developed a dynamic image filter based on IE by studying the former work and try to prevent erotic information' s spread efficiently based on analyzing and comprehending the image content. The results show that the system works well. Some key techniques are included, such as skin detection, face detection, object area segmentation, image features extraction, the design of classifier and the implement of filter based on browser.In period of skin detection, we adopted an effective skin color model by studying the former work and based on this we carried out skin detection through simple and statistical texture character. Then we get skin mask from original images by the Skin Model. Based on skin mask image, a method based on independent component analysis is used to detect faces in image. The face detection model is divided into two parts, extraction of independent component, and face detection with FastICA. Based on the work above, we adopted classified method based on Support Vector Machines(SVMs). SVMs have advantage in resolving the small sample, non-linear, high dimension pattern recognition and classification problem, so we apply it to erotic image filtering system to classify the images .The paper utilizes the technique of BHO(Browser Helper Object) to carry out the erotic web page filter. BHO import code to the IE address space to process the images in web page by the classifier' s judge . BHO is a COM(Comopent object Model) object borned by Internet Explorer...
Keywords/Search Tags:Image Filtering, Skin Model, Face detection, Support Vector Machines (SVMs), Browser Helper Object(BHO)
PDF Full Text Request
Related items