| Research Abstract |
To extract object regions from images, the methods using region-based active contour model (ACM) have been proposed. By controlling ACM with the statistical characteristics of the image properties in regions, these methods effect robust region extraction. However, the existing methods require redundant processing and cannot adapt to complex scene images.
To achieve efficient and accurate region extraction, we proposed a new method for controlling region-based ACM. In the proposed method, to prevent redundant processing, a definite support area is set along an object boundary, and ACM is controlled by the statistical characteristics of the image properties in this area. Furthermore, to improve the accuracy of region extraction, the proposed method performs clustering on the support area and introduces this result into ACM control. Consequently, the image properties in a necessary and sufficient area can be effectively reflected on the region extraction process by the proposed method. Through the region extraction experiments, we confirmed the effectiveness of the proposed method.
In the existing region-based ACMs, since it is difficult to determine the appropriate search areas, the search areas have been set heuristically. For this difficulty, the proposed method can determine the appropriate sizes and locations of the search areas in parallel to an object boundary. However, in perpendicular to an object boundary, the proposed method cannot determine the appropriate search area sizes. To deal with this problem, in addition to the characteristics of the image properties, a priori knowledge of an object shape should be introduced efficiently into ACM control.
|
