Rethinking Image Color Correction, Validation and Testing

Abstract

Digital image capture normally includes a color-correction step that transforms detector signals into corresponding image pixel values. For digital cameras and scanners, we usually base the color-correction operation on captured images of reference color charts. Measures of object color-capture are included in recent imaging guidelines for cultural heritage institutions. Several methods have been adopted as standard practice, with the aim of reducing image-capture variation. During the evaluation of the goodness of object-to-image color-encoding, there is normally a validation step. This involves comparing the original target colorimetry to that of the predicted colors and calculating summary color-difference metrics for the population of target samples. While this is an instinctive and common approach we believe it needs to be revisited. The current summary statistics for evaluating color capture goodness can be misleading for the color-content at hand. Additionally, reporting color error measurements for the same colors that were used to develop the color-correction is effectively 'teaching to the test' when evaluating digital capture color performance. We discuss strategies for selecting validation colors based on generic and specific use cases along with examples.