Method for Combining Image Segmentation Maps on the Basis of Information Redundancy and Variation of Information Minimization
Dmitry M. Murashov · Optoelectronics Instrumentation and Data Processing · 2022
Abstract A new two-level method based on the minimization of a dual-objective quality functional for combining image segmentation maps is proposed. The functional is formed as a weighed sum of information redundancy and variation of information values calculated from an original image and a combined segmentation map. Such a measure results in the image segmentation providing a compromise between the conditions of minimizing the quantity of selected informationally significant segments and the information dissimilarity between an original image and a resulting segmentation. The proposed method makes it possible to improve the segmentation result as compared to the method for combining segments by the criterion of an information redundancy minimum.