IMAGE DETECTIVE: EFFICIENT IMAGE RETRIEVAL SYSTEM

Abstract

With the wide - spread use of image retrieval in various areas such as crime investigation, medical diagnosis, intellectual property rights, etc, today’s need is to enhance the image retrieval process. In our research, we are combining Text Based Image Retrieval (TBIR) method with Content Based Image Retrieval (CBIR) method to enhance image retrieval. The base of CBIR is to extract different image features, such as Color, Shape and Texture. To improve the accuracy, we are using combination of most efficient feature extraction algorithms. We are using RGB to Lab conversion for color feature extraction, Modified Canny edge detection algorithm with variable sigma for shape feature extraction, Framelet transform method for texture feature extraction. For improving the speed of image retrieval process using TBIR, we are implementing automatic annotation technique. Images are annotated automatically without human intervention. It improves speed. Approximately one to two thousand images are stored in the database. Features are extracted from these images and stored into the database. Query images are processed in the similar way and similarity matching between query and database images is done through Hybrid Graph method. For that purpose, we have to generate image to image graph from extracted feature vectors and image to tag graph from database. Combining both these graphs, we get the Hybrid graph. Thus, the process of image retrieval is becoming efficient in both terms accuracy and time. Also, user can give input in terms of query image or textual query or sketch. This improves human – friendliness of this system.