Surgical Smoke Dehazing and Color Reconstruction

Abstract

Smoke in minimally invasive robotic surgeries appears during the cauterization of vessels and dissection of tissues. The smoke obstructs the field of view of the surgeon in real-time surgeries. Use of energy instruments for smoke removal have their own limitations. Many dehazing or desmoking algorithms reported induce color distortions and artifacts, which can mislead the surgeons. In this paper, a workflow for smoke dehazing and color restoration is proposed. We automate the process of air-light estimation using smoke features which is the key parameter for dehazing. To reduce color artifacts a color normalization is applied to the dehazed image using the detected non-smoky reference image. The results of the proposed approach shows a reduction in color artifacts and outperforms other methods reported in the literature.