Global and Local Graph-Based Difference Image Enhancement for Change Detection

Change detection (CD) is an important research topic in remote sensing, which has been applied many fields. In the paper, we focus on post-processing of difference images (DIs), i.e., how to further improve quality a DI after initial obtained. The importance DIs for CD problems cannot be overstated, however few methods have investigated so far re-processing their acquisition. order quality, propose global and local graph-based DI-enhancement method (GLGDE) specifically problems; this plug-and-play that can both homogeneous heterogeneous CD. GLGDE first segments multi-temporal into superpixels with same boundaries then constructs two graphs as vertices: one feature graph characterizes association between similarity relationships connected vertices changing states DI, other spatial exploits change information contextual DI. Based these graphs, model built, constrains enhanced smooth graphs. Therefore, proposed not only but also correct it. By solving minimization model, obtain improved experimental results comparisons different tasks six real datasets demonstrate effectiveness method.