CUNSB-RFIE: Context-aware Unpaired Neural Schrödinger Bridge in Retinal Fundus Image Enhancement
Xuanzhao Dong, Vamsi Krishna Vasa, Wenhui Zhu, Peijie Qiu, Xiwen Chen, Yi Su, Yujian Xiong, Zhangsihao Yang, Yanxi Chen, Yalin Wang
Abstract
Retinal fundus photography faces challenges in image quality due to systemic and human-related factors. This work introduces the CUNSB-RFIE framework, leveraging Schrödinger Bridge and Dynamic Snake Convolution to enhance retinal images, effectively preserving fine structures like blood vessels and outperforming existing methods.
Figures (click on each for a larger version):
