A New Approach for CBIR based on Color Coding-MTSD with Pattern Recognition Neural Network

Content Based Image Retrieval (CBIR) is the method which uses visual data to find the images from a huge database on the foundation of user’s input in conditions of query image and usually some effective features for color string, HSV histogram, edge orientation and intensity map are extracted. This study proposes an picture characteristic illustration manner centered on Color Coding-Multi-Trend Structure Descriptor (CMTSD) and Pattern Recognition Neural Network (PRNN). We overview the outcome on Corel dataset which includes a thousand and 5000 pictures. And the experimental results based on precision, accuracy and feature extraction (FE) time. The proposed algorithm is in similarity with MTSD procedure and it is better than MTSD approach. The distance between two matrixes is calculated using different similarity measures, namely, L1, Euclidean distance (ED), ChebyChev, Hamming and Jaccard distance. Neural network (NN) is used for categorization of query picture as per training database. At first NN is trained about the image features in the database. The training is done by using pattern network algorithm. This trained database is used for classification of the query image. According to retrieved image class further CMTSD based CBIR is used for retrieving similar images.