A Novel Quadtree-Salient Image Decomposition for Color Doppler Flow Imaging

Color Doppler flow imaging analysis is one of those difficult problems in medical image decomposition fields. To address the key problem, we propose a new bottom-up segmentation method - quadtree-salient decomposition which is suitable for machine vision applications in complex environments. In our proposed method, a LMS perception space converting processing will firstly be carried on and the three independent channel could be extracted as the cornerstone of decomposition. After this, an integrated visual salient model which includes Itti, GBVS and image signature models is proposed. At last, we give the detailed quadtree-salient decomposition algorithm which is fitted for human visual system. Experimental results on the real CDFI dataset demonstrate that the proposed method has advantages over LoG, Canny and the classical quadtree decomposition.