Novel gabor-PHOG features for object and scene image classification

Abstract

A new Gabor-PHOG (GPHOG) descriptor is first introduced in this paper for image feature extraction by concatenating the Pyramid of Histograms of Oriented Gradients (PHOG) of all the local Gabor filtered images. Next, a comparative assessment of the classification performance of the GPHOG descriptor is made in six different color spaces, namely the RGB, HSV, YCbCr, oRGB, DCS and YIQ color spaces, to propose the novel YIQ-GPHOG and the YCbCr-GPHOG feature vectors that perform well on different object and scene image categories. Third, a novel Fused Color GPHOG (FC-GPHOG) feature is presented by integrating the PCA features of the six color GPHOG descriptors for object and scene image classification, with applications to image search and retrieval. Finally, the Enhanced Fisher Model (EFM) is applied for discriminatory feature extraction and the nearest neighbor classification rule is used for image classification. The effectiveness of the proposed feature vectors for image classification is evaluated using two grand challenge datasets, namely the Caltech 256 dataset and the MIT Scene dataset. © 2012 Springer-Verlag Berlin Heidelberg.