A Novel Approach for CBIR using Color Strings with Multi-Fusion Feature Method

In this paper, proposed a novel approach for a very popular image retrieval technique namely ‘Content Based Image Retrieval’. This method works by combining all the low level features like color, texture, shape and skin color for creating a hybrid approach for efficient and accurate retrieval. A new method Color Strings is used for color feature extraction, DWT is utilized for extricating texture component from the image, for extracting shape feature edge histogram with five directional orientation is used and lastly for detecting skin color firstly a color balancing algorithm is applied and after that skin color is detected based on YCbCr color space. For analyzing the performance of the method a dataset is constructed Corel-1250 which consist of five categories of images namely Sports, Flowers, Fruits, Tools and Faces. Euclidean Distance, Relative Standard Deviation and Cityblock Distance metrics are used for similarity comparison. The performance is measured utilizing precision, recall and accuracy values.