Performance evaluation of CBIR system based on object detection and evolutionary computation

This paper discusses the performance evaluation of the Content Based Image Retrieval (CBIR) system using the optimality in selection of feature vector elements. The performance of the CBIR system may be improved by appropriate analysis of the image. Image analysis is still facing problems related to the detection of the objects. In spite of the noticeable achievements using the part based model, the improvement in detection of objects still demands more attention. The algorithm proposed here for Content Based Image Retrieval is characterized by a LBPHOG based object descriptor and an evolutionary computation technique for the optimum features selection. A popular and widely used method for segmentation based on Unsupervised Curve evolution is deployed here. The optimum selection of the feature vector elements is controlled by the fitness function of the Simple GA used here. This right selection of feature vector elements improves the efficiency of the algorithm. The Algorithm is tested on the Berkeley database that contains the images those are characterized by the low depth. The experimental results show that the proposed algorithm achieves promising results for this data base.