Accelerated Path Tracing With GAN and Matrix Completion

Denoising-based techniques have recently been shown to be effective for accelerating path tracing rendering methods. However, there remains a problem which is input images need the minimum necessary samples number in order ensure quality of output. In this paper, we propose new accelerated approach with generative adversarial networks(GAN) and matrix completion. Unlike methods based on denoising neural network, randomly render part pixels image, are much less than other Next, utilize trained GAN pre-complete initializing missing pixels. Because accuracy fast-convergence GAN, our pre-completion results more accurate Then, according pre-completion, present pre-completed as low-rank make use completion recovers values accurately even high details. To improve efficiency solving completion, modified original weighted nuclear norm minimization parameter adjustment(PAWNNM) strategy. The result shows better visual quality, texture details convergence state-of-the-art acceleration methods, especially network.