Multi-resolution local binary patterns for image classification

This paper presents a novel method to extract image features for image classification. The extracted feature named multi-resolution local binary pattern (MR-LBP) is based on the local binary pattern (LBP) feature. The MR-LBP feature is highly distinctive by making use of multi-resolution patterns to obtain more descriptive information. The experiments results demonstrate the proposed MR-LBP feature is robust to image rotation, illumination changes and image noises. We also describe a descriptor called MR-LBP descriptor to using the features for image classification. Through experiments, our proposed approach performs favorably compared with the most well-known SIFT descriptor in two benchmark dataset. What's more, the proposed descriptor is computation simpler than the SIFT descriptor.