Browsing by Author "Varshney, Amitabh"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Accelerated Volume Rendering with Volume Guided Neural Denoising
    (The Eurographics Association, 2023) Jabbireddy, Susmija; Li, Shuo; Meng, Xiaoxu; Terrill, Judith E.; Varshney, Amitabh; Hoellt, Thomas; Aigner, Wolfgang; Wang, Bei
    Monte Carlo path tracing techniques create stunning visualizations of volumetric data. However, a large number of computationally expensive light paths are required for each sample to produce a smooth and noise-free image, trading performance for quality. High-quality interactive volume rendering is valuable in various fields, especially education, communication, and clinical diagnosis. To accelerate the rendering process, we combine learning-based denoising techniques with direct volumetric rendering. Our approach uses additional volumetric features that improve the performance of the denoiser in the post-processing stage. We show that our method significantly improves the quality of Monte Carlo volume-rendered images for various datasets through qualitative and quantitative evaluation. Our results show that we can achieve volume rendering quality comparable to the state-of-the-art at a significantly faster rate using only one sample path per pixel.
  • Thumbnail Image
    Item
    Real-time Monte Carlo Denoising with the Neural Bilateral Grid
    (The Eurographics Association, 2020) Meng, Xiaoxu; Zheng, Quan; Varshney, Amitabh; Singh, Gurprit; Zwicker, Matthias; Dachsbacher, Carsten and Pharr, Matt
    Real-time denoising for Monte Carlo rendering remains a critical challenge with regard to the demanding requirements of both high fidelity and low computation time. In this paper, we propose a novel and practical deep learning approach to robustly denoise Monte Carlo images rendered at sampling rates as low as a single sample per pixel (1-spp). This causes severe noise, and previous techniques strongly compromise final quality to maintain real-time denoising speed. We develop an efficient convolutional neural network architecture to learn to denoise noisy inputs in a data-dependent bilateral space. Our neural network learns to generate a guide image for first splatting noisy data into the grid, and then slicing it to read out the denoised data. To seamlessly integrate bilateral grids into our trainable denoising pipeline, we leverage a differentiable bilateral grid, called neural bilateral grid, which enables end-to-end training. In addition, we also show how we can further improve denoising quality using a hierarchy of multi-scale bilateral grids. Our experimental results demonstrate that this approach can robustly denoise 1-spp noisy input images at real-time frame rates (a few milliseconds per frame). At such low sampling rates, our approach outperforms state-of-the-art techniques based on kernel prediction networks both in terms of quality and speed, and it leads to significantly improved quality compared to the state-of-the-art feature regression technique.