A Survey of Computational Color Constancy: Approaches and Emerging Applications in Medical Imaging

As the evening sets in and natural light fades to artificial light, it is remarkable how the colors around us still appear nearly the same. This phenomenon, known as color constancy, allows us to perceive relatively stable colors even as lighting conditions change. In the field of computer vision, replicating this human ability is the key to performing tasks such as recognizing objects and understanding scenes. This review examines two main approaches to tackle the challenge of computational color constancy: traditional static methods that rely on a set of assumptions about lighting and scene conditions, and newer learning-based methods that use models trained to estimate lighting directly from images. We look at how these techniques have evolved, from foundational methods like the gray-world assumption, which work within certain limitations, to modern deep learning techniques that can interpret lighting in more dynamic ways. The datasets, metrics, and reported results of the surveyed studies are consolidated into two comparative tables, providing a structured overview of the methods discussed. Together, these comparisons offer a broad perspective on how computational color constancy has developed and where it is headed, with particular emphasis on its applications in medical imaging.