HIK Based Image Classification Using Saliency Based Multiscale Representation

IK based image classification method is proposed which efficiently classifies the images with less background. HIK kernel is used with saliency driven multiscale fusion. The generated scale space in general preserves or even enhances semantically important structures such as edges, lines, or flow like structures in the foreground, and smoothes clutter in the background. The image is classified using multiscale information fusion based on the original image, the image at the final scale at which the diffusion process converges, and the image at a midscale. The proposed method emphasizes the foreground features, which are important for image classification. The background image regions, whether considered as contexts of the foreground or noise to the foreground, can be globally handled by fusing information from different scales.HIK kernel is used for classification, to handle the cases of incorrect detection of saliency.HIK based visual code book generation algorithm is used for visual vocabulary creation, in order to classify the images with less background effectively. Experiments were conducted in 17 Oxford Flowers, 102 Oxford Flowers and Caltech 101 datasets. The proposed method has an accuracy of 81.3%. Keywords— multiscale fusion,visual vocabulary, classification,saliency, HIK.