Multi-Light Image Collections (MLICs) are stacks of photos of a scene acquired with a fixed viewpoint and a varying surface illumination that provides large amounts of visual and geometric information. Over the last decades, a wide variety of methods have been devised to extract information from MLICs and have shown its use in different application domains to support daily activities. In this thesis, we present methods that leverage a MLICs for surface analysis and visualization. First, we provide background information: acquisition setup, light calibration and application areas where MLICs have been successfully used for the research of daily analysis work. Following, we discuss the use of MLIC for surface visualization and analysis and available tools used to support the analysis. Here, we discuss methods that strive to support the direct exploration of the captured MLIC, methods that generate relightable models from MLIC, non-photorealistic visualization methods that rely on MLIC, methods that estimate normal map from MLIC and we point out visualization tools used to do MLIC analysis. In chapter 3 we propose novel benchmark datasets (RealRTI, SynthRTI and SynthPS) that can be used to evaluate algorithms that rely on MLIC and discusses available benchmark for validation of photometric algorithms that can be also used to validate other MLIC-based algorithms. In chapter 4, we evaluate the performance of different photometric stereo algorithms using SynthPS for cultural heritage applications. RealRTI and SynthRTI have been used to evaluate the performance of (Neural)RTI method. Then, in chapter 5, we present a neural network-based RTI method, aka NeuralRTI, a framework for pixel-based encoding and relighting of RTI data. In this method using a simple autoencoder architecture, we show that it is possible to obtain a highly compressed representation that better preserves the original information and provides increased quality of virtual images relighted from novel directions, particularly in the case of challenging glossy materials. Finally, in chapter 6, we present a method for the detection of crack on the surface of paintings from multi-light image acquisitions and that can be used as well on single images and conclude our presentation.

Surface analysis and visualization from multi-light image collections

Tinsae Gebrechristos Dulecha
2021-01-01

Abstract

Multi-Light Image Collections (MLICs) are stacks of photos of a scene acquired with a fixed viewpoint and a varying surface illumination that provides large amounts of visual and geometric information. Over the last decades, a wide variety of methods have been devised to extract information from MLICs and have shown its use in different application domains to support daily activities. In this thesis, we present methods that leverage a MLICs for surface analysis and visualization. First, we provide background information: acquisition setup, light calibration and application areas where MLICs have been successfully used for the research of daily analysis work. Following, we discuss the use of MLIC for surface visualization and analysis and available tools used to support the analysis. Here, we discuss methods that strive to support the direct exploration of the captured MLIC, methods that generate relightable models from MLIC, non-photorealistic visualization methods that rely on MLIC, methods that estimate normal map from MLIC and we point out visualization tools used to do MLIC analysis. In chapter 3 we propose novel benchmark datasets (RealRTI, SynthRTI and SynthPS) that can be used to evaluate algorithms that rely on MLIC and discusses available benchmark for validation of photometric algorithms that can be also used to validate other MLIC-based algorithms. In chapter 4, we evaluate the performance of different photometric stereo algorithms using SynthPS for cultural heritage applications. RealRTI and SynthRTI have been used to evaluate the performance of (Neural)RTI method. Then, in chapter 5, we present a neural network-based RTI method, aka NeuralRTI, a framework for pixel-based encoding and relighting of RTI data. In this method using a simple autoencoder architecture, we show that it is possible to obtain a highly compressed representation that better preserves the original information and provides increased quality of virtual images relighted from novel directions, particularly in the case of challenging glossy materials. Finally, in chapter 6, we present a method for the detection of crack on the surface of paintings from multi-light image acquisitions and that can be used as well on single images and conclude our presentation.
Multi-Light Image Collections , RTI, NeuralRTI
File in questo prodotto:
File Dimensione Formato  
Surface_analysis_from_Multi_Light_Image_Collections.pdf

accesso aperto

Tipologia: Tesi di dottorato
Licenza: Dominio pubblico
Dimensione 41.57 MB
Formato Adobe PDF
41.57 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11562/1043402
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact