Complementary horizontal visibility patches: from texture classification to remote sensing image classification

Texture information has been widely used in remote sensing image classification to improve the accuracy of feature recognition. This paper presents new applications of complex network theory and tools to texture classification and remote sensing image classification to recognize texture patterns. The proposed complementary horizontal visibility patches (CHVP) algorithm naturally extends the horizontal visibility graph (HVG) algorithm. The initial model of the complex network is first constructed by taking the image pixel points as network nodes. Then, the initial network model dynamically evolved using the CHVP method to extract network features at various stages of evolution and describe image structural features. Finally, it was tested on standard Kylberg, UMD, and SIRI-WHU data sets, which have general accuracies of 100. 00%, 95.70%, and 70. 5%, respectively, compared to the local binary pattern (LBP), adaptive center pixel selection (ACPS), simple siamese (SimSiam), and local ternary pattern (LTP), among others. This new feature discovery method has important potential implications in texture modeling.