The need for efficient Image Retrieval has increased tremendously in many application areas and in addition to it the present day images are extremely varied with lot of information. Hence the problem of Image Retrieval has grown further complex. Implementing CBIR based on single feature like color, texture or shape does not produce satisfactory results. In our proposed approach, the retrieval is carried out on the user selected region i.e., ROI (Region - Of - Interest) followed by evaluating the low level features. These multi-features are fused based on the similarity score and the fitness function is evaluated by the Genetic Algorithm (GA). In GA the weights of similarity score are optimally assigned. The Corel databases of 1000 images are considered in which image retrieval is done for actual and ROI image. The performance is evaluated by the parameters recall rate and precision rate. From the obtained results, it is evident that our proposed approach outperforms traditional methods.
展开▼
