Image Quality Assessment of Hyperspectral and Conventional Imaging for Cultural Heritage Artifacts

Abstract

Preserving historical artifacts is crucial as they carry cultural significance beyond their artistic value and offer insight into their creation for future generations. Digitalization of cultural heritage artifacts preserves them by enabling high-resolution digital images for enhanced accessibility, broader, and improved conservation.It also offers adaptability for various applications and reduces the need for physical handling, minimizing the risks of damage and long-term deterioration. However, the digital representation of cultural heritage faces challenges in the adaptation of technology, economic sustainability, standardization, and quality control. To ensure accuracy, completeness, fidelity, and legibility compared to the original object, integrating quality assessment is essential during the digitization life cycle of cultural heritage. Therefore, this thesis seeks to investigate and identify essential parameters for characterizing and evaluating the behavior and content of digitized artifacts in the domain of cultural heritage. The research presented in this thesis covers both RGB imaging and spectral imaging techniques, focusing on two important cultural heritage artifacts; paintings and microfiche. For artwork such as paintings, it is essential to identify and classify pigments accurately. It helps curators and conservators determine the historical context, authenticity, and proper conservation methods. Conventional RGB imaging systems are valuable in various applications, including cultural heritage. Nevertheless, their limited spectral range makes them impractical when additional information is needed, such as material identifications, composition analysis, detection of hidden features, etc. Spectral imaging, such as multispectral and hyperspectral imaging, overcomes these limitations by capturing much broader spectral information. In this thesis, hyperspectral imaging is used to analyze paintings. The spectral quality of the imaging system was evaluated on its accuracy in identifying pigments for classification purposes. Various spectral metrics based on supervised algorithms and machine learning models were used and analyzed for pigment classification. improper handling, and poor storage conditions. Consequently, digitization becomes necessary. Microfiche relies on specialized devices for direct human eye reading, primarily available in select archives or libraries. However, over time, these devices can suffer from various problems that affect their usability. Given the importance of preserving and accessing microfiche materials, in this thesis, we investigated alternative options for microfiche digitization. The process of digitizing microfiche involves magnification, which can introduce various factors that can impact the overall reproduction quality, such as noise, distortion, and artifacts. Additionally, these materials can suffer from poor legibility and clarity because of their reduced size, potential degradation over time, or even limitations of the scanning device. To address this, objective and subjective image quality assessments are considered in this thesis to assess the quality of the digitized microfiche. The objective of this Ph.D. research is to evaluate the quality of digitization of cultural heritage artifacts. Through a compilation of articles, this dissertation contributes to overall research objectives and seeks to provide a comprehensive understanding of the research subject to assess the quality of digitized cultural heritage artifacts using conventional and hyperspectral imaging technologies. The analysis and practical recommendations of these articles discussed in this thesis are coherent and provide valuable resources for researchers and practitioners involved in the preservation and documentation of cultural heritage and those involved in imaging, and provide valuable insights into potential directions for future research.