Abstract
The lack of depth perception in 2D image-guided surgical navigation makes surgical procedures error-prone, especially when precise operation is required. Integral photography (IP), along with animated integral videography (IV), is an autostereoscopic 3D imaging and display technique that can produce true 3D images with stereo and motion parallax. The computer-generated IV rendering process for imaging 3D surface models or volumetric data simulates the original IV process using computer graphics, so it requires more computations than the normal rendering process and cannot achieve real-time imaging using only a CPU. In this paper, we first discuss the essence of IP from the viewpoint of a 4D light field. We then introduce the method we developed for real-time 3D imaging of both anatomical models and medical volumes, which uses consumer-level GPU-based IV, and present our flexible hybrid framework for GPU-based IV rendering. Finally, we describe the evaluation we performed using 3D imaging examples showing that our GPU implementation is at least 97 times faster than a CPU implementation and at most 400 times faster.
