Pattern and Color Descriptors for Retrieval of Images

For the retrieval of images, CBIR (Content Based Image Retrieval) technique is used for extraction of features from the images. Various approaches exist for the feature extraction from the images in the database, but they lack in accuracy. To solve this - a hybrid technique is proposed in this paper for efficient feature extraction of images. Local Binary texture patterns such as LBP (Local Binary Pattern) and CS-LBP(Center-Symmetric Local Binary Pattern) are used with color descriptors such as Color Histogram and Color correlogram to form a hybrid approach to make retrieval more accurate and efficient. Euclidean Distance classifier is used for the similarity matching between the query image and the images stored in the database. The proposed technique has an accuracy of 0.86 with precision and recall factors of 0.92 and 0.93 respectively which is fairly large then the techniques used so far in isolation.