Font Size: a A A

Research On Semantic Retrieval Of Face

Posted on:2014-03-26Degree:MasterType:Thesis
Country:ChinaCandidate:G L YangFull Text:PDF
GTID:2208330422988325Subject:Computer software and theory
Abstract/Summary:
With the network technology, information technology and the rapid developmentof computer technology, face image resources will become more and more. How tocreate fast and efficient face image retrieval mechanism, has become the field ofimage processing problems to be solved. In this paper, based on semantic features ofthe human face looks retrieval method, we propose a bottom face image to theadvanced features of the semantic mapping method; To avoid cumbersome betweenthe image and the image matching process, we propose a dual weights searchalgorithm. To sum up this research work is mainly reflected in the following aspects:First, the use of color-based face detection technology, the use of Gaussian mixturemodel image from a sample library automatically locate the face position; Next, aseparate organ Facial Modeling using a simple model to simulate the complexgeometry of the human face organs, according to the characteristics of differentorgans in different organs of the human face segmentation method, will face majororgans from the human face images are automatically segmented; then, through thefacial organs geometric feature point extraction, and set varying amounts of facialorgans characteristic parameters, according to the range of these parameters foradvanced mapping semantic word; Finally, these high-level semantic words to expressa different combination of face images stored to face the semantic database and designa human face looks based on semantic retrieval system, system input face advancedsemantic description, and then search through the search algorithm face semanticdatabase, retrieve a certain number of face images in order to achieve fast andaccurate semantic retrieval purposes.166people establish a human face databaseretrieval experiment results of the previous eight out correctly retrieved160people,accounting for96.39%of total experiment. Experimental results show that thealgorithm has higher retrieval efficiency. Use this method to establish the semanticface image retrieval system, with fast, efficient and practical features.
Keywords/Search Tags:Face looks retrieval, the parameters of the geometric characteristicsof the semantic description, doublet search algorithm
Related items