Time sequence gradient domain rendering method and system based on neural network

• Main Authors: XU YANNING, WANG LU, MA MINGCONG
• Format: Patent
• Language: Chinese; English
• Published: 15.03.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention provides a time sequence gradient domain rendering method and system based on a neural network, and relates to the field of graphics realistic rendering, and the specific scheme is as follows: scene data to be rendered comprises continuous image frames, and each frame is processed as follows: a renderer tracks an intersection point of light and a scene in the current frame, and the image frame is rendered; recording channel information of the intersection points in the tracking process, wherein the channel information comprises a color channel, a gradient channel, an auxiliary feature channel and a motion vector channel of the current frame; constructing a time sequence mapping channel of the current frame by using the color channel and the motion vector channel, and adding the time sequence mapping channel into channel information of the current frame; inputting the pre-processed channel information into a pre-trained neural network, and taking an output reconstructed image as a rendering resul