Image Segmentation Using Region Merging Combined with a Multi-class Spectral Method

In this paper we propose an image segmentation algorithm that combines region merging with spectral-based techniques. An initial partitioning of the image into primitive regions is produced by applying a region merging approach which produces a chunk graph that takes in attention the image gradient magnitude. This initial partition is the input to a computationally efficient region segmentation process that produces the final segmentation. The latter process uses a multi-class partition that minimizes the normalized cut value for the region graph. We have efficiently applied the proposed approach with good visual and objective segmentation quality results.