Content-Based Image Retrieval (CBIR) has received a comprehensive attention from researchers due to the quickly growing and the diffusion of image databases. Despite the huge research efforts consumed for CBIR, the completely promising results have not yet been presented. In this paper, a novel weighted multi-feature voting technique is proposed which incorporates various types of low-level visual features such as texture, shape and color in retrieval process. The color feature is described by color histogram and hierarchical annular histogram whereas shape feature is described by edge histogram and edge direction histogram while texture feature is described by gabor filter and co- occurrence matrix. Each feature has certain weight computed based on its precision to reflect its importance in retrieval procedure. Furthermore, different distance measures are implemented to get the highest precision of each feature. The results indicate that by applying multi-features and multi-distance measures, the obtained retrieval system outperforms other existing methods with accuracy 89.5% for Wang database, 91.5% for Caltech101 database and 89% for UW database.