Relevance Feedback through the Generation of Trees for Image Retrieval Based on Multitexton Histogram

The Content-based image retrieval (CBIR) systems and their application in different areas of development, are current research topics, however the semantic gap between low-level image features and high-level semantic concepts handled by the user, is one of the main problems in the image retrieval. On the other hand, the relevance feedback has been used on many CBIR systems such as an effective solution to reduce the semantic gap. For that reason this paper proposes a method of relevance feedback based on the generation of trees and Histogram Multitexton descriptor. This method has been compared with the conventional RF algorithms "Query vector modification", and show significant improvements in terms of effectiveness in the image retrieval. Also the dimensionality of the Histogram Multitexton descriptor has been tested and with the first 64 dimensions increase its effectiveness which permit to reduce the computational processing time.