Measurement texture geometric feature reconstruction method based on photometric stereo

Abstract

The invention discloses a measurement texture geometric feature reconstruction method based on photometric stereo. The method is characterized in that utilizing a realistic texture measuring device in a camera light source array mode to collect BTF data under different angle light sources and viewing angles; with measurement texture multi-view angle information being combined, optimizing a photometric stereo method and reconstructing normal information of the measurement texture; and reconstructing depth information of the measurement texture through a least square method in a measurement gradient field according to the normal information. More accurate normal information can be calculated when the input image sequence is influenced by non-Lamertian diffuse reflection parts of highlight and shadow and the like in the stage of calculating shadow information of the surface of the texture, and in the stage of calculating the depth information, influence of noise of the measurement gradient field on the depth is reduced through filtering.

Description

technical field [0001] The invention relates to a method for reconstructing geometric features of measured materials based on photometric stereo vision. Background technique [0002] In computer graphics, realistic rendering of complex materials such as cloth fabrics and skins has always been a great challenge. The difficulty with this type of problem is that the highly realistic rendering of a material relies on its geometric modeling as well as its material modeling. Geometric modeling describes the shape characteristics of objects, which is to construct the macroscopic geometric structure of the visible concave-convex features on the surface of the object, while the traditional macroscopic geometric model is not capable of expressing the material; The micro-geometry of the specular distribution. Different objects of the same geometric model, due to differences in surface geometry, material, density, color and other attributes, will present complex and rich optical pheno...

AI Technical Summary

Problems solved by technology

However, since the reflection model is assumed to be an ideal Lambert reflection model, in fact, there are non-Lambert abnormal parts in the input image sequence, which are easily affected by factors such as highlights, shadows, and noises, making the results of surface reconstruction deviate, and there are low-frequency noise

And photometric stereo vision has strict requirements on measuring equipment, which is easy to introduce system errors

Images