Face Image Retrieval Based On Semantic Features Of Eyes

With digital network technology and the rapid development of information technology,multimedia, visualization face image resources become more and more. The digital imageprocessing technology, face pattern recognition, computer vision processing technology andtraditional database image retrieval technology combine to establish an efficient face imageretrieval system; an image is now a popular field of information technology researchquestions. In recent years, semantic image retrieval is the human face human face imageretrieval research and development of a new trend. The main component of human eyes, face,has a relatively significant and stable characteristics of the human face semantic imageretrieval have important significance. Semantic features based on eye face image retrievalonly improve the human face image retrieval speed and reduces the retrieval algorithmcomplexity.Establish facial feature database and optimize database search algorithm is semanticimage retrieval research faces two main directions. Face feature extraction to more accurateface image retrieval accuracy rate is higher, so this paper to establish semantic facial featuredatabase, optimize the face image region segmentation method, and refined the classificationof facial shape characteristics. First, the use of adaptive color mapping based on facedetection method to get the face region, and then use based on facial skin color eye areaimage automatic segmentation method belt region obtained by dividing the human eye image;shape of the eye to create your own description of the shape of the human eye characteristicgeometric model and defines four categories of eye shape features and characteristics of leftand right eyes separately; eye position according to the image obtained by dividing theeyebrow area and shape eyebrows defines five characteristics; Finally extracted to the humaneye and eyebrow semantic encoding feature to create a feature database. In order to enhancethe efficiency of semantic image search, the paper face image retrieval algorithm is also madefurther optimized, the first principal component analysis-PCA on the characteristic parametersof dimension reduction, reuse, not equidistant retrieval coefficient factor calculation methodset match obtain characteristic values closest to the human face, analyze search results using asimple method to modify the feedback coefficient factor characteristic parameters to achieveoptimal results.