Color correction algorithms for digital cameras

The image recorded by a digital camera mainly depends on three factors: the physical content of the scene, the illumination incident on the scene, and the characteristics of the camera. This leads to a problem for many applications where the main interest is in the color rendition accuracy of the scene acquired. It is known that the color reproduction accuracy of a digital imaging acquisition device is a key factor to the overall perceived image quality, and that there are mainly two modules responsible for it: the former is the illuminant estimation and correction module, the latter is the color matrix transformation. These two modules together form what may be called the color correction pipeline. This thesis has the objective to design and test new and more robust modules for the color correction pipeline, studying and exploiting the existing crosstalks in order to obtain a higher color reproduction accuracy. The first module considered is the illuminant estimation and correction one; many illuminant estimation solutions have been proposed in the last few years, although it is known that the problem addressed is actually ill-posed as its solution lack uniqueness or stability. To cope with this problem, different solutions usually exploit some assumptions about the statistical properties of the expected illuminants and/or of the object reflectances in the scene. In the last few years two research areas that are important in the context of improving the performance of color constancy algorithms have been highlighted: making additional measurements at the time of image capture (i.e. using more color pixel information), and algorithm combining (i.e. using two or more estimations of the illuminants). In this thesis a third hypothesis is investigated: the use of low level visual information to improve illuminant estimation. The second module considered is the transformation of the camera-dependent RGB image data into a standard RGB color space. This transformation, usually called color correction matrix or color matrixing, is needed because the spectral sensitivity functions of the sensor color channels rarely match those of the desired output color space (usually sRGB). The color correction matrix transformation is usually optimized assuming that the illuminant in the scene has been successfully estimated and compensated for. Both the illuminant estimation process and the color correction matrix concur in the formation of the overall perceived image quality. The two processes have always been studied separately, thus ignoring the interactions between them. In this thesis the interactions between the illuminant estimation process and the color correction matrix in the formation of the overall color accuracy are investigated, especially when the white point estimation is imperfect. How the color correction transform amplifies the illuminant estimation errors is also investigated. Furthermore, it is shown that it is possible to incorporate knowledge about the illuminant estimation behavior in the optimization of the color correction matrix to alleviate the error amplification. It is demonstrated that a fixed device chromatic response characterization, which is often adopted, is not able to produce good color accuracy in most situations. New strategies to improve color accuracy under both perfect and imperfect white point estimation are proposed, which clearly suggest that adaptive color transformations have to be preferred in order to improve the color accuracy.