A multi-spectral image database and its application to image rendering across illumination

摘要

We present a database of full spatial and spectral resolution images of everyday objects captured with a novel multi-spectral device: the applied spectral imaging Spectracube camera. We summarise the operation of the device and present a study of its spectral accuracy before demonstrating the usefulness of the data set by using it to evaluate different white point correction transforms: transforms which map RGBs under one illuminant to a corresponding set of RGBs under a different light. We find that while the best transforms can often give results close to the results obtained using spectral techniques, the results are both image and illuminant dependent and in a number of cases all transforms tested introduce significant errors.