Emergence of Internet, as well as digital image acquisition technology, has increased the usage of rich visual information such as images and videos. It has become the integral and essential part of everyone’s life. Since image production has become easy and economical, images and videos are used extensively on the Internet so retrieving relevant images from large ever-growing image dataset has become a challenge. To combat this problem, one of the popular image retrieval approaches is Content-Based Image Retrieval (CBIR) which utilizes the visual features of the image i.e. color, texture, geometric (shape) and spatial information to retrieve visually similar images according to the given query image. This paper aims to exploit multiple features of an image i.e. color, geometric and texture with the Back-propagation Feedforward Neural Network (BFNN) for classification. Feature selection method is also exercised to focus on important features of an image and ignoring redundant and inappropriate information. The results have shown that CBIR technique with multi-feature classification by BFNN yields better precision and recall as compared to other state-of-the-art techniques of CBIR that uses single or other hybrid combination of features of an image.