A high compression ratio bidirectional reflection material compression representation model, method and device

Abstract

The invention discloses a bidirectional reflective material compression representation model with a high compression ratio, which includes an encoder and a decoder. The encoder is used to compress bidirectional reflective material information into a probability distribution on a low-dimensional representation space, and its input is a direction An encoded data pair composed of information and corresponding reflection brightness information, the output is a probability distribution on a low-dimensional representation space; the decoder is used to decode and generate reflection brightness information according to a fixed-length vector on a low-dimensional representation space, and its input is direction information and the decoded data pair composed of the fixed-length vector extracted from the probability distribution, and the output is the reflection brightness information. This model can realize the compression of the high-dimensional original reflection material into a low-dimensional representation space, which reduces the memory usage of the original reflection material, and the low-dimensional representation space can also ensure the recovery accuracy. Also disclosed is a high compression ratio two-way reflective material compression representation method and device.

Description

technical field [0001] The invention relates to the technical fields of graphics drawing, material representation and deep learning, and in particular to a high compression ratio two-way reflection material compression representation model, method and device. Background technique [0002] In the field of material representation, the collection and representation of complex materials will be of great help to the graphic application of drawing high-precision materials. It has a wide range of needs and applications in the film industry, game production, virtual reality and other fields. In particular, the modeling application of high-precision models is becoming more and more extensive, which puts forward higher requirements for high-quality material design. However, directly using the collected complex material data has problems such as large storage space and few query times. In addition, the collected complex material data cannot be modified intuitively, so that more requir...

AI Technical Summary

Problems solved by technology

[0006] First of all, the method based on parametric model fitting is difficult to fit, and the performance capability is limited by the designed parametric model. At present, there is no parametric model that can cover the real material, and many real collected material features will be lost.

[0007] In addition, based on the method of tensor data, the compression ability of the main feature extraction technology is weak, and the loss of information is serious; the fitting non-parametric function model technology can only compress the material information, cannot represent the space for the material construction, and cannot judge the relationship between materials. similarity, and the inability to visually modify the material

Images