A segmentation method using Otsu and fuzzy k-Means for stereovision matching in hemispherical images from forest environments

In this paper we describe a novel pixel-based strategy of segmentation and stereovision matching for obtaining disparity maps from hemispherical images captured with fish-eye lenses from forest environments. At a first segmentation stage, the method identifies textures of interest to be either matched or discarded and extracts six attributes of each pixel as features. This is achieved by applying both Otsu and fuzzy k-Means methods. It is a combination of strategies appropriately sequenced to automate the process and facilitate the matching. At a second stage, a stereovision matching process is designed based on the application of three stereovision matching constraints: epipolar, similarity, and uniqueness. The epipolar guides the process. The similarity and uniqueness are mapped through a decision making strategy based on a majority voting criterion. The main finding of this paper is the combination of strategies in the both stages. The method is compared against the usage of simple features and some existing similarity matching strategies using also combination.