Weighted Constrained Hue-Plane Preserving Camera Characterization.

Abstract

Color correction relates device dependent sensor responses (RGB) to device independent color values (XYZ). Here we present a new approach to Hue-plane Preserving Color Correction (HPPCC) using weighted constrained 3 × 3 matrices. Hue-plane preservation was introduced in [1] in conjunction with an HPPCC method. That method maps using a finite number of local white point preserving 3 × 3 matrices, each of which operates in a hue-angle delimited subregion of device space defined by the white and two adjacent chromatic training set colors. However, that formulation does not leave room for optimization or continuity beyond C0 in the transitions between the subregions. To remedy that our new method uses hue-angle specific weighted matrixing: given a device RGB from which a device hue-angle is derived, a corresponding transformation matrix is found as the normalized weighted sum of all precalculated constrained white point and training color preserving matrices. Each weight is calculated as a power function of the minimum difference between the device and the training color hue-angle. The weighting function provides local influence to the matrices that are in close hue-angle proximity to the device color. The power of the function is optimized for global accuracy. We call this Hue-plane Preserving Color Correction by Weighted Constrained Matrixing HPPCC-WCM 1 1. Experiments performed using different input spectra show that our method consistently improves on both stability and accuracy compared to state of the art methods.