SHAFT (SAT & HUE Adaptive Fine Tuning), a new automated solution for target-based color correction

Target-based 'color correction' is today a popular technique to have a faithful reproduction of color in many field where the digital photography is used: portraiture, fashion, furniture design, interior design and Cultural Heritage (CH). This technique, establishing the color relationship according to a set of color patches with available pre-measured spectral or colorimetric data, allows to have color corrected images using a limited set of parameters and subtle variation of color according with well-defined effects. Following a growing success, essentially due to its flexibility and easy to use, in the last years several color correction target-based solutions appeared both commercial and open source. These solutions follows basically two different methods to calculate the transformation between measured CIEXYZ values and captured RGB values: linear transformations, or least-squares polynomial regressions. In many cases, a simple linear transformation is sufficient to map device-dependent and device-independent spaces with adequate performance. Linear approaches are extensively employed for 'color correction' as they preserve two key properties directly related to the camera sensors linear response to the light sources: scalability and hue planes. Despite above listed benefits, linear correction may produce significant errors. To allow better estimations an efficient option is the polynomial color correction. For fixed calibration settings, a polynomial regression can strongly reduce the mapping error allowing significant improvements to color correction. However, the use of high degree data expansions can result in unstable (rank deficient). Another major problem in the use of the polynomial regression consists in its not scale-independent feature (i.e., data scaled as result of changes in the scene radiance or exposure). This shift can be significant in several cases and this is a significant problem in outdoor captured images. A third solution, could be found in the Adobe Camera Raw (ACR) calibration scripts coming from Bruce Fraser's calibration procedure for successive approximations. This accurate solution, however presents the problem to be completely grounded on the Adobe Photoshop software and a further problem consisting in the stop of the development in 2010. In this paper we present a new ACR-like solution completely written in MATLAB able to be used alone or coupled with a polynomial regression, introducing several optimizations and enhancement to the original algorithm. The so-named new SHAFT differs from the original technique for the number and types of tests done along the processing and for the algorithm used to find the best variation from the original values of the selected parameters (exposure, contrast, white balance, hue and saturation on each RGB channel). Tests of the new solution in many field related to the CH and typical problems of each class of algorithms are illustrated showing surprising results.