Content-based image retrieval (CBIR) has most important research area in the last couple of years. Image retrieval is the system used for searching out the desired features of a query image. The main purpose of CBIR method is to obtain the accurate and fast results. In this paper, an Query image is retrieved from the database with high accuracy. This going through various steps namely, Pre-processing, feature extraction, Optimization and matching. In pre-processing noise is removed by using filtering and color conversion is performed on the query image. For extracting features SURF (Speedup Robust Feature) descriptor is used. It is the enhanced form of SIFT (Scale Invariant Feature transform) and more speedy than it. For optimization Genetic algorithm (GA) is used which improves the features of the extracted image. For matching Euclidian formula is used. It measures the similarity between the extracted feature and the database image. The simulation is carried out in MATLAB simulator tool with an accuracy of approximately 98 %.