A Baseline Statistical Method for Robust User-Assisted Multiple Segmentation

Recently, several image segmentation methods that welcome and leverage different types of user assistance have been developed. In these methods, the inputs can be provided by drawing bounding boxes over objects, scribbles or planting seeds help to differentiate between boundaries interactively refining missegmented regions. Due variety in amounts inputs, relative assessment becomes difficult. As a possible solution, we propose simple yet effective, statistical method handle utilize input amounts. The proposed is based on robust hypothesis testing, specifically DGL test, implemented with time complexity linear number pixels quadratic Therefore, it suitable used as baseline for quick benchmarking assessing performance improvements user-assisted algorithms. We provide mathematical analysis operation method, discuss its capabilities limitations, design guidelines present simulations validate operation.