Consistent validation of gray-level thresholding image segmentation algorithms based on machine learning classifiers

We propose a Machine Learning approach for Image Validation (MaLIV) to rank the performances of two or more outputs obtained from different gray-level thresholding image segmentation algorithms. MaLIV utilizes machine learning classifiers to rank automatically the outputs of different segmentation algorithms accounting for both the computational complexity of the validation experiment and for the robustness of its results. The proposed method resorts to subsampling to find Fisher consistent estimates of validity measures obtained from a sample of pixels of extremely-reduced size. To this purpose, subsampling is combined with three alternative approaches: learning curves, asymptotic regression and convergence in probability. Results of experiments involving the validation of five images segmented through thirteen different algorithms are presented.