Sunita Saha has successfully defended her thesis!

Sunita Saha has successfully defended her thesis!

We are very proud to announce that March 22nd, Sunita Saha defended her thesis titled “Automated Identification of Changes from Cultural Heritage Surfaces”. In April, the degree was officially awarded by Warsaw University of Technology, Poland. Congratulations, dr. Sunita Saha!

Abstract of Saha’s thesis

This thesis aims to develop segmentation methods to detect, quantify and visualize the changes on the surface of cultural heritage (CH) objects over time. The development incorporates automated and interactive multi-modal data processing considering 3D and Reflectance Transformation Imaging (RTI) data.

The first part of the work (section 3.1) is centered on developing the change-based segmentation method on 3D data. The technical advances for the development of the method were to quantify and visualize the changes on the same surface by comparing two-time data instances. The development pipeline considered several crucial factors like noise, sampling density, cross-time alignment error, and adaption of the thresholds for solving the bias in the results by each factor. The advancement of the method was tested with the existing 3D model comparison methods.

The work’s second part (section 4.1) focuses on developing a robust supervised segmentation of RTI appearance attributes. The Polynomial Texture Mapping (PTM) and Hemispherical Harmonics (HSH) descriptors were considered for supervising the reconstruction coefficients to a definitive class. During the development of the method, several variance parameters of the data acquisition system were tested, and the solution was adapted through data normalization. After supervision of the data, a discriminant model called trained data was created and then applied to test the sample data. The development was also compared with the existing image segmentation method to claim the advancement of the developed method.

Following the segmentation methods, a user-friendly color map visualization was generated, representing the outcome of the method for the ease of everyday users improving digital documentation. Finally, the application of the developments was tested considering several CH objects.

Digital solutions to different conservation questions like condition assessment, monitoring, and before and after conservation treatment (cleaning, restoration of damages and deformation, etc.) are addressed in the thesis. The correctness of the proposed methods is based on the conservators’ opinion considering it as the ground truth of this work. For each considered case study for this work, the results were verified by a group of conservation and restoration experts.

Sunita Saha’s thesis is available here.