Endoscopic Image Exposure Correction

We began by selecting normally exposed images from several prominent endoscopic image repositories, including Kvasir, CVC-ColonDB, CVC-ClinicDB, ETIS-LaribPolypDB, and CVC-300. To simulate exposure anomalies reflective of real-world variations, we utilize the LECARM model, which allows us to apply a random exposure range of (-1,1). This generates a diverse spectrum of exposure anomalies that closely resemble those encountered in clinical settings.

本研究伊始，我们从包括 Kvasir、CVC-ColonDB、CVC-ClinicDB、ETIS-LaribPolypDB 以及 CVC-300 在内的多个杰出内窥镜影像库中挑选了常规曝光的图像。为了模拟现实世界中的曝光异常，我们采用了 LECARM 模型，该模型能够允许我们应用 (-1,1) 的随机曝光范围。这生成了一系列多样化的曝光异常，其与临床环境中遭遇的异常高度相似。