NOVEL APPROACHES IN COLOR ASSESSMENT: FROM POINT-WISE COLORIMETRY TO FILM RESTORATION

Colorimetry is the science which studies the quantification, reproduction and management of color from its optical, chemical, physical, but also physiological and perceptual point of view. Today, colorimetry has been standardized and color management can count on a robust workflow in point-wise conditions, thus when a color stimulus is considered under specific constraints (e.g., standard illumination and observation). As a consequence, the rigorous application of standard colorimetry is insufficient to reproduce and manage color in real conditions with non-uniform illumination and with complex spatial arrangements. This work of research is organized into three threads of work, which follow the standard pipeline of color signal evaluation: color assessment, color acquisition and processing, and quality assessment. In the first Part of this study, the assumptions and constraints of standard colorimetry are recalled, together with specific comments on their limits in non-standard and practical applications. The open problems and misuses of colorimetry discussed from the theoretical point of view are supported by preliminary tests and experiments, aiming at highlighting the limits of the application of standard colorimetry in cultural heritage field. In the second Part of this work, the standard digitization workflow is analyzed, and the issues and effects which could affect the acquisition of contrast and tones in imaging systems are discussed. In particular, the limits of hyperspectral imaging have been tested and examined, in order to assess the main source of noise in the acquisition and to evaluate the reliability of the acquired data. Subsequently, an implementation of the digitization protocol for film restoration is proposed and different methods to perform image enhancement and processing have been examined. For this aim, the family of Spatial Color Algorithms (SCAs), derived from Retinex, has been found successful for cultural heritage applications and a novel approach to film restoration is proposed. This specific family of algorithms, enhance colors according to the spatial distribution of pixel values in the scene, thus include the visual spatial mechanisms in color computation, overcoming the main constraints of standard colorimetry. Nevertheless, SCAs present high computational costs. To address this problem, in this Thesis a new speed-up algorithm is presented in order to allow the enhancement of video streams. Regarding quality assessment thread, an overview of image quality metrics is presented, and different metrics and measures have been examined and tested, to define which metrics could be the most effective and usable for cultural heritage applications. In this context, a new framework of image quality measures has been presented and applied on historical images and videos. The field of cultural heritage has been particularly suitable to underline the limits in the application of standard colorimetry, because, in this context, color is never observed as isolated phenomenon, but always inside a spatial arrangement. In this work, film digital restoration has been the main field of application and different approaches and alternative methods have been proposed to overcome the limits of standard color analysis. Anyway, the highlighted limits and solutions are applicable to film restoration, as well as to many other fields, where color sensation cannot be limited to point-wise colorimetry.