Texture Based Superpixel Segmentation Algorithm for Hyperspectral Image Classification

Abstract

Superpixel segmentation is a potential preprocessing tool that can simplify an image by creating semantic sub-regions of the image. This is quite effective in reducing the misclassification of isolated pixels. Image segmentation using superpixels has been successfully applied in the computer vision field. Recently, superpixel segmentation is being adopted into HSI processing domain also, due to its superior properties. In this work, for the accurate classification of the Hyperspectral Image, a texture-based superpixel segmentation algorithm is proposed. Local Binary Pattern (LBP) and Gabor filters are first incorporated to extract the local and global image texture information. Next, Simple Linear Iterative Clustering (SLIC) superpixel segmentation algorithm is applied over the extracted texture features and a superpixel segmentation map is obtained. Finally, majority voting strategy is performed between the superpixel segmentation map and the pixel-wise classification map to acquire the final classification map. For validating the effectiveness of the proposed algorithm, experiments were conducted on four popular HSI datasets, namely: Indian Pines, Pavia University, Houston 2013, and Houston 2018 datasets. Superior classification performance was observed by the proposed method in comparison to other state-of-the-art methods.