Probabilistic Color Image Classifier Based on Volumetric Robust Features
Need of more sophisticated methods to handle color images becomes higher due to the usage, size and volume of images. To retrieve and index the color images there must be a proper and efficient indexing and classification method to reduce the processing time, false indexing and increase the efficiency of classification and grouping. We propose a new probabilistic model for the classification of color images using volumetric robust features which represents the color and intensity values of an region. The image has been split into number of images using box methods to generate integral image. The generated integral image is used to compute the interest point and the interest point represent the volumetric feature of an integral image. With the set of interest points computed for a source image, we compute the probability value of other set of interest points trained for each class to come up with the higher probability to identify the class of the input image. The proposed method has higher efficiency and evaluated with 2000 images as data set where 70 % has been used for training and 30% as test set.
