Statistical Quantification of Texture Visual Features for Pattern Recognition by Analyzing Pre- and Post-Multispectral Landsat Satellite Imagery
Amit Kumar Shakya, Ayushman Ramola, Anurag Vidyarthi
Abstract
This paper investigates the performance of visual texture features based on the gray-level co-occurrence matrix to analyze multispectral remotely sensed Landsat images. Different case studies related to urbanization, flood, and drought are investigated in this research work. The changing land use/land cover pattern caused by urbanization, floods, and droughts is examined through quantitative assessment. Texture visual features, i.e., correlation, contrast, angular second moment or energy, and homogeneity, are derived from the gray-level co-occurrence matrix. These features are used to develop a pattern for the changing texture of land use/land cover. Human visual perception of smoothness and coarseness is related to the texture features and is later used to describe the texture features’ changing behavior. The quantitative assessment of texture features in terms of smoothness and coarseness establishes a novel pattern between pre- and postimages of urbanization, flood, and drought.