三篇透明物体重构参考英文文献

Title:Depth Reconstruction of Translucent Objects from a Single Time-of-Flight Camera using Deep Residual Networks
Author(s)：Seongjong Song and Hyunjung Shim
Source:Computer Vision – ACCV 2018,March26,2019
Main ideas:It is the first attempt to solving the 3D reconstruction of translucent objects using deep neural networks.
Main findings or ideas:
1、The paper adopt the deep residual networks for modeling the ToF depth distortion caused by translucency.
2、To fully utilize both the local and semantic information ofobjects, multi-scale patches are used to predict the depth value.
Conclusion:This paper present a deep residual network architecture for recovering depth distortion from translucent object, using a single time-of-flight (ToF) depth camera.The quantitative and qualitative evaluations over the competitor clearly demonstrate the superiority of the model; The author report higher accuracy and show the robust performance for handling various object poses and optical properties. In addition, the experimental validation of their proposals justifies positive effects. By showing the robustness of the model across various levels of input noise, the author highlight that the model can be a practical solution for real applications. The paper propose the first approach to recovering the 3-D translucent object using deep neural networks.

Title:Inspection of extremely slight aesthetic defects in a polymeric polarizer using the edge of light between black and white stripes
Author(s)：Yuan-long Deng, Shao-peng Xu, Hao-quan Chen, Zhan-heng Liang, Ci-long Yu
Source:Polymer Testing，February5,2018
Main ideas:This study proposed a novel and automated inspection method for detecting ESTADs by employing the edge of light in structured light illumination, and this was found to drastically enhance the image contrast.
Main findings or ideas:
1、The paper propose a novel method of using edges of light for inspecting ESTADs. ESTADs can be detected at the edge of light regions under structured light illumination.
2、On the basis of a previous study, an optic model was also developed to explain the imaging enhancement mechanism of the ESTADs detection method.
3、Finally,The paper successfully detected ESTADs using four-step scans and demonstrated the feasibility of this technique for the detection of ESTADs in industry.
Conclusion:An optical model was built for simulation to investigate the imaging enhancement mechanism, and the lighting principle was also presented to describe the imaging characteristics of a defect on the edge of light—a set of luminous pixels near the dark stripe and set of dark pixels near the luminous stripe.

Title:Recovering Translucent Objects using a Single Time-of-Flight Depth Camera
Author(s)：Hyunjung Shim, Member, IEEE, Seungkyu Lee, Member, IEEE
Source:IEEE Transactions on Circuits and Systems for Video Technology, May,2016
Main ideas: This paper described the development of a unified 3D data acquisition framework that reconstructs translucent objects using a single commercial Time-of-Flight (ToF) camera.
Main findings or ideas:
1、The paper developed a new distance representation that interprets the depth distortion induced as a result of translucency.
2、By analyzing ToF depth sensing principles,the paper constructed a distance model governed by the level of translucency, foreground depth, and background depth.
Conclusion:In this paper, the author propose a new framework for recovering a translucent object using a single ToF camera. The framework is reproducible, operable by an ordinary user, and practical because it is not necessary to modify camera hardware or firmware. In future studies, the experiment will extend this framework for multiple depth cameras and acquisition of multi-view depth maps that can used to simultaneously estimate the foreground, background and a translucent parameter.