Segmentation of LSCM Images based on Multi-channel Information Fusion

The two channels of LSCM, the fluorescent light channel and the visible light channel, provide us with different modality images that contain special information respectively. In this paper we propose a new and integrated approach to segment images in the fluorescent light channel of LSCM, which have rather low SNR and can not provide sufficiently high intensity gradient at the boundary. Our approach, rather than relying on information of the velocity field alone, also includes statistical information of images in the visible light channel which provide subtle information of vessel structures. Information is described by corresponding image force. The approach is tested on LSCM images and experimental results show that it can segment low SNR vasculature structures automatically. Comparison is made between C-V model and the new approach, we find that the latter has better performance and can provide vasculature delineation with higher quality since information of both channels is utilized.