Automatic Segmentation of Fluorescence Endoscopic Images Based on the Weighted Otsu Method
Xin Yang, Natalia A. Obukhova, Ivan I. Ivanov · 2022 Conference of Russian Young Researchers in Electrical and Electronic Engineering (ElConRus) · 2022
The multi-spectral fluorescence endoscope system uses a near-infrared camera system to image the fluorescence signal. By fusing the near-infrared image with the white light image of the same visual field, realizing the intraoperative fluorescence navigation. The basis of this system is the precise segmentation of the fluorescent region in the near-infrared image. In this paper we proposed an automatic segmentation method for the fluorescence region of near-infrared fluorescence endoscopy images. This method is based on the Otsu method, which uses the multi-step weighted Otsu method to segment the image. The segmentation quality is estimated by the misclassification error of the image. Compared with the traditional Otsu method, this method improves the efficiency of segmentation of fluorescent regions: the misclassification error of the proposed method is 4%, while Otsu is 16%.