Real-time endoscopy haze removal: a synthetical method

In endoscopic surgeries, the smoke and haze generated by temperature difference and electrosurgical knife usage degrades the surgical view. While many researches on natural single image haze removal being proposed, few works are done for video haze removal of endoscopy view. In this paper, we proposed a synthetical method for endoscopy haze removal combining a Refined Dark Channel Prior (RDCP), Spatial-Temporal Markov Random Field (STMRF), Color Attenuation Prior (CAP) and parameter self-adaption. Qualitative and quantitative experiment results outperformed some traditional and deep learning based methods. Our proposed method is an effective and practical one.