Region segmentation for environment understanding using spatial frequency
Dae-Nyeon Kim, Kang-Hyun Jo · Proceedings. The 8th Russian-Korean International Symposium on Science and Technology, 2004. KORUS 2004. · 2005
This paper presents a recognition method for region segmentation in the image. We segment the regions from the image previously defined as objects. To define the objects, we used the feature which contains color, texture, shape and geometrical information. We select such image information according to the characteristics of objects. The recognition method adopts to understand outdoor environment by extracting representative objects. We added the additional information such as shape and texture of the objects to recognize contrast to the conventional approach. Such a method shows relatively good result especially in the region where the spatial frequency varies according to the objects and vicinity. This method is very useful when the outdoor environment image is used as the robot can navigate.