A new semantic-based color image retrieval
Lu Zhou · Journal of Liaoning Normal University · 2009
The performance of content-based image retrieval(CBIR) systems is largely limited by the between the low-level feature and high-level concept.In this paper,a new semantic-based color image retrieval scheme is proposed,which not only takes into consideration the important image edge information and human visual system,but also utilizes the support vector regression(SVR) theory.Firstly,the important image edge is extracted by using canny detection operator and the low-level texture feature and color feature are computed.Then,the high-level object is determined by mapping the low-level feature using SVR.Secondly,the high-level emotion is determined by mapping the low-level line direction using SVR.Thirdly,the high-level color can be obtained by using the main color of the quantized color image.Finally,image retrieval is implemented by using the above high-level object semantic,emotion and color semantic.Experimental results show that the proposed image retrieval are effective in characterizing image high-level and can provide sound and robust image retrieval performance,which has strong significance for reducing the semantic gap between the visual feature and concept.