Parallel Interactive Ray-Space Rendering Approach and Implementation

A real time interactive ray-space rendering approach is proposed. Multiview video(MW) becomes more and more important in the Held of image processing, and FTV as an application of MW based on ray-space is required to be interactive and real time. Both of the state-of-the-art rendering techniques ray-space and Lightfield have to handle enormous data, thus the complexity of rendering multiviews in real time is very high. The common approaches based on sequential processing or multi-core processing just can render 4 views in real time at most. The proposal can render 16 views which can be controlled by the view control interactive signals in real time. The parallel architecture divided the 16 views into 4 view groups (GOP). The GOPs are being processed simultaneously, thus the rendering time is decreased to 1.08 ms/frame. The simulation shows that the proposal has the same rendering quality as the rendering on software, we implement the approach on FPGA platform, and also test the FPGA design on Modelsim. The implementation results show that the logic resource cost on FPGA platform is only 139 slices.