Multi-kernel denoising autoencoder for accelerating particle-based rendering

Morii, Masahiro*; Sakamoto, Naohisa*; Kawamura, Takuma 

Particle-based rendering, which does not require view-based sorting, is effective for visualization of large simulation data. However, in ensemble averaging, the increase in the number of ensembles for high image quality may interfere with interactive operations. To solve this problem, we propose a multiple kernel denoising autoencoder that improves the image quality of low ensemble images. In experiments, we tested the effectiveness of our method on several test data sets.