対象物体に依存しない領域間の情報に基づいた領域合併による物体抽出手法

概要

論文の詳細を見る
We propose a method for segmenting a color image and extracting a region of object in a scene using information which is not specific to a target object but common to a projection of an object in a space. The proposed method is based on split-and-merge, and at first an image is segmented into small regions by clustering of colors. Then the regions are merged with four features of information about configuration, connection and area ratio between two adjacent regions. This merging enables us to extract an object as one region even if the surface of the object is textured with a variety of reflectances; this isn't considered by conventional segmentation methods. We show the algorithm of segmentation and extraction, and demonstrate experimental results and comparisons with some other methods.
社団法人映像情報メディア学会の論文
2001-04-20