A robust and consistent stack generalized ensemble-learning framework for image segmentation

Abstract In the present study, we aim to propose an effective and robust ensemble-learning approach with stacked generalization for image segmentation. Initially, input images are processed feature extraction edge detection using Gabor filter Canny algorithms, respectively; our main goal is determine most descriptions. Subsequently, applied stacking technique, which generally built two learning levels. The first level composed of algorithms that give good results in literature, namely: LightGBM (Light Gradient Boosting Machine) SVM (support vector machine). second meta-model use a predictor model takes base-level predictions improve accuracy final prediction. process, Extreme (XGBoost); it as sub-models’ outputs better classify each pixel Today, several research works exist literature different machine algorithms; fact, instead trying find single efficient optimal learner, ensemble-based techniques take advantage basic model; they integrate their obtain more consistent reliable learner. result obtained from models individuals proposed compared set evaluation measures quality such IoU, DSC, CC, SSIM, SAM, UQI. comparison showed model. Thus, have made some recent deep learning-based unsupervised segmentation methods. coherent terms precision, robustness, consistency.