Disparity Map Prediction from Stereo Laparoscopic Images using a Parallel Deep Convolutional Neural Network

Bálint Antal

Abstract: One of the main computational challenges in supporting minimally invasive surgery techniques is the efficient 3d reconstruction of stereo endoscopic or laparosocopic images. In this paper, a Convolutional Neural Network based approach is presented, which does not require any prior knowledge on the image acquisition technique. We have evaluated the approach on a publicly available dataset and compared to a previous deep neural network approach. The evaluation showed that the approach outperformed the previous method.

TL;DR: A CNN-based approach to map stereo image pacthes to disparity maps.

Conflicts: inf.unideb.hu

Keywords: Computer vision, Deep learning, Applications