Perceptual grouping of natural images for CBIR

In current developments of the CBIR system, it is desirable to design a computer program that can index the content of an image automatically into a set of perceptually significant components. Results from existing image segmentation techniques are not sufficient to represent the content of an image. Further grouping is required to produce more meaningful segmentation. This paper describes our approach to implement Gestalt principles for region grouping. Results from image segmentation are further grouped into regions representing major components of image content.