Multimodal image classification using inverted local patterns

Multimodality during capturing images suffers from significant contrast variation between the images of the same scene. Due to this large variation, existing image classification and retrieval algorithms are not performing well for multimodal images. So, to solve this problem of classifying multimodal images, we have proposed a modality invariant descriptor based on a local pattern description method named Local Binary Pattern (LBP). The quantitative results show that the proposed descriptor outperforms not only other state of the art modality invariant descriptors but also famous LBP variants in terms of classification accuracy.