Hair point cloud neural rendering method with stable time domain

Abstract

The invention discloses a time-domain-stable hair point cloud neural rendering method, and provides a deep stripping and time-domain stability enhancement network for solving the problem that an existing neural rendering network based on point cloud cannot render high-quality hair with time-domain stability, and through hierarchical projection of an input point cloud model, the time-domain stability of the hair is enhanced. Feature information of different layers is obtained, and results are fused to adapt to the semitransparent feature of hair; and inputting the trained result into the time domain stability enhancement network, obtaining the dependency relationship between the current frame and the previous frames by using the re-projection of the point clouds between the adjacent frames, and generating the final result of the current frame, thereby ensuring the time domain stability of the training result. According to the method, the result of global illumination is inferred through the information of local illumination, and the rendering quality and the rendering speed can be ensured at the same time. The defect that a traditional method uses global illumination algorithms such as ray tracing to render hair is very time-consuming is overcome.

Description

technical field [0001] The invention relates to the field of computer graphics and the field of neural rendering, in particular to a time-domain stable hair point cloud neural rendering method. Background technique [0002] Hair plays a vital role in the photorealistic representation of computer-generated characters and animations. In the film industry, for example, in the movies "Godzilla vs. King Kong" and "Zootopia", there are abundant hairs around the animal bodies. Immersive and immersive experience. Also in the gaming industry, the high-quality hair that a large number of characters in "Sekiro" has is very important for gamers to get a better gaming experience page. [0003] Therefore, how to generate high-quality hair is a very important issue. Traditional hair is generally expressed using a shape based on patches or curves, and is rendered by rasterization or ray tracing, which may cause problems of poor shading effect or slow rendering speed. And if you use an o...

AI Technical Summary

Problems solved by technology

Traditional hair is generally expressed based on patches or curves, and rendered by rasterization or ray tracing, which may cause problems such as poor shading effect or slow rendering speed

And if you use an offline renderer to render a frame of high-quality hair, it usually takes half an hour to an hour, which is very expensive

Images