Content based image retrieval using joint descriptors

Abstract

Content based image retrieval helps manipulators to retrieve pertinent images based on their contents. A consistent content-based feature extraction technique is required to meritoriously extract most of the information from the images. These important elements include color, texture, intensity or shape of the object inside image. Various descriptors required for extracting global and local features. It's easy to describe message in the image linguistically but it's difficult to do the same for machine applications. Combined descriptors like canny edge detector, Histogram and Discrete Wavelet Transform are used. Histogram extracts global features, wavelet transform extracts texture feature in four orientations(±45°, ±90° and ±180°). Histogram, Canny edge detector jointly used to form joint feature descriptor. Result shows that histogram as descriptor outperforms result with an efficiency of 75% with feature vector size of 1×6.