Generic visual categorization using composite Gabor and moment features

In this paper an effective method to recognize objects from different categories of images which suffer from illumination, variability in shape, occlusion and clutter, based on a combination of spatial and spectral features called new composite features is presented. In domain of object recognition, it is often to classify objects from images that make only limited part of the image. Hence to identify local features and certain region of images, patches are extracted over the interest points detected from the original image using Wavelet based interest point detector. Gabor features and Moment features are computed separately for every patch and classified using SVM classifier. In addition to this, Gabor features are combined with Moment features, so-called new composite features are computed for every patch and its performance is compared with the independent features. The observations revealed that composite features outperform the independent features with less error rate. The experimental evaluation is done using the Caltech database.