Improvement of Colorization Realism via the Structure Tensor

Colorization is a colour manipulation mechanism employing user-assisted colour hints for changing greyscale images into coloured ones. Several colorization algorithms have been constructed, and many of these methods are able to produce appropriately colorized images given a surprisingly sparse set of hints supplied by the user. But these colour images may not in fact look realistic. Moreover, the contrast in the colorized image may not match the gradient perceived in the original greyscale image. We argue that it is this departure from the original gradient that contributes to the un-real appearance in some colorizations. To correct this, we make use of the Di Zenzo gradient of a colour image derived from the structure tensor, and adjust the colorized image such that the Di Zenzo definition of the maximum-contrast gradient agrees with the gradient in the original grey image. We present a heuristic method to this end and guided by this approach devise an optimization-based method. Our gradient projection tends to result in more natural-looking images in the resulting adjusted colorization. To explore the proposed method we utilize minimalist sets of colour hints and find in particular that “hotspots” of un-realistic colour are subdued into regions of more realistic colour. This paper is not aimed at introducing a new basic colorization but instead our method is meant to make any colorization look more realistic; we demonstrate that this is the case for several different basic methods. In fact, we even find that a very simplistic colorization algorithm can be used provided the projection proposed here is then used to make the colorization more realistic looking.