Image segmentation with multi-feature fusion in compressed domain based on region-based graph

Image segmentation plays a significant role in image processing and scientific research. In this paper, we develop a novel approach, which provides effective and robust performances for image segmentation based on the region-based (block-based) graph instead of pixel-based graph. The modified Discrete Cosine Transform (DCT) is applied to obtain the Square Block Structures (DCT-SBS) of the image in the compressed domain together with the coefficients, due to its low memory requirement and high processing efficiency on extracting the block feature. A novel weight computation approach focusing on multi-feature fusion from the location, texture and RGB-color information is employed to efficiently obtain weights between the DCT-SBS. The energy function is redesigned to meet the region-based requirement and can be easily transformed into the traditional Normalized cuts (Ncuts). The proposed image segmentation algorithm is applied to the salient region detection database and Corel1000 database. The performance results are compared with the state-of-the-art segmentation algorithms. Experimental results clearly show that our method outperforms other algorithms, and demonstrate good segmentation precision and high efficiency.