Layout Estimation for Layered Ink of 3D Printer to Reproduce the Desired Line Spread Function of Skin using Simulated Data

Abstract

In this article, the authors propose a method to estimate the ink layer layout for a three-dimensional (3D) printer. This enables 3D printed skin to be produced with the desired translucency, which they represent as line spread function (LSF). A deep neural network in an encoder–decoder model is used for the estimation. It was previously reported that machine learning is an effective way to formulate the complex relationship between optical properties such as LSF and the ink layer layout in a 3D printer. However, although 3D printers are more widespread, the printing process is still time-consuming. Hence, it may be difficult to collect enough data to train a neural network sufficiently. Therefore, in this research, they prepare the training data, which is the correspondence between an LSF and the ink layer layout in a 3D printer, via computer simulation. They use a method to simulate the subsurface scattering of light for multilayered media. The deep neural network was trained with the simulated data and evaluated using a CG skin object. The result shows that their proposed method can estimate an appropriate ink layer layout that closely reproduces the target color and translucency.

Perceptual Translucency in 3D Printing Using Surface Texture

“…In this method, an encoder-decoder deep neural network is used to determine the optimal multilayer layout; however, the training data are created using patches actually fabricated by a 3D printer, which is costly. Nagasawa et al thus developed a method of reproducing multilayer patches on a computer by simulating light scattering using a Monte Carlo method to generate training data [24]. Another approach reproduces the appearance of realistic skin by extracting a concentration map of pigment components from an image of human skin taken with an RGB camera and fabricating a multilayered structure that imitates actual skin [25].…”

Fabrication of a Human Skin Mockup with a Multilayered Concentration Map of Pigment Components Using a UV Printer

“…Second and third clear layers with two, four, or six constituent layers were prepared, and nine samples were created with different combinations of these thicknesses, as shown in Table 1 . The scattering characteristics of Mimaki’s clear inks have been investigated in the references [ 25 ], and it has been shown that a small amount of scattering occurs. Therefore, changing the thickness of the clear layer may cause inconsistency with the data of 900 color patches; however, in this study, we ignored this effect as it is very small.…”

“…One study used multilayered skin modeling performed with a neural network-based method to estimate the optimal layer structure layout for reproducing human skin with arbitrary skin color and transparency [ 24 ]. Furthermore, a study has succeeded in reducing the fabrication cost by using simulation data [ 25 ]. However, these studies assumed that the skin is a uniform surface, whereas in reality, its color varies spatially.…”