Image data is now commonly represented in any of a myriad of file formats, from proprietary raw formats through standards like JPEG or PNG. In addition to recording pixel values, most of these formats contain metadata that defines how pixel values translate into colors and brightnesses. Unfortunately, when non-trivial computations are used to transform pixel values, it is often necessary that both the metadata and file format used be changed. For example, using a program called parsek, this tool aligns images and produces a “raw” super resolution result which has problematic shifts in color and tonality. The goal of the current work has been to understand what is causing these shifts and how they can be avoided. Toward that goal, the color and tonal metadata information representations in a variety of file formats is reviewed. The effectiveness of preserving appearance when file format is changed is then reviewed. A reference color and tonal rendering is obtained by examining specific metadata and preview renderings embedded in several file formats.
Coleman Earlywine, Henry Dietz, "Preservation of Color Metadata Across Transformations of Pixel Data" in Electronic Imaging, 2025, pp 204-1 - 204-7, https://doi.org/10.2352/EI.2025.37.11.HVEI-204