Perceptually Optimal Reproduction of Color Images considering Device Limits

在拟议的项目中，应考虑到输出设备的限制(打印机、显示器和投影仪)来研究多光谱、高动态范围图像的感知最佳再现。特别令人感兴趣的是有限的同色异谱和光谱色域，以及这些设备的低动态范围。以前的方法只关注一个设备限制(色域映射、光谱色域映射和HDR色调映射)。然而，为了考虑到人类视觉系统的各种外观现象，需要一种统一的方法。为此，应修改现有的图像外观模型，以便将图像映射到感知上一致的外观相关空间。该项目的一个基本目标是开发基于人类空间和颜色视觉的该空间内的图像差异测量。对于多光谱图像，利用这种图像差度量可以构造一系列受设备限制的耦合变分问题。解决方案是操作员，它将图像转换为设备极限，与原始图像的知觉差异最小。应开发新的数值方法来解决这些变分问题。应特别重视它们的数值复杂性、收敛速度和稳健性，以便在合理的时间内也处理高分辨率多光谱图像。将设计并实现一个原型多光谱再现系统，以显示该方法的可行性、其在计算性能方面的适用性，并通过与现有方法的心理物理比较实验来验证其最优性。

In the proposed project the perceptually optimal reproduction of multispectral, high dynamic range images shall be investigated considering output-device limits (printers, displays and projectors). Of special interest are the limited metameric and spectral gamut as well as the low dynamic range of these devices. Previous approaches focused only on one single device limit (gamut mapping, spectral gamut mapping and HDR tone mapping). To allow for various appearance phenomena of the human visual system, however, a unified approach is necessary. For this purpose existing image appearance models shall be modified in order to map the image into a perceptually uniform appearance correlate space. One basic goal of this project is the development of an image difference measure within this space based on human spatial and color vision. For a multispectral image a sequence of coupled variational problems subject to device limits can be constructed by means of this image difference measure. The solution is an operator, which transforms the image into the device limits with minimal perceptual disagreement to the original. New numerical methods shall be developed to solve these variational problems. Particular importance shall be attached to their numerical complexity, convergence speed and robustness in order to handle also high resolution multispectral images in reasonable time. A prototypical multispectral reproduction system shall be designed and realized to show the feasibility of this approach, its applicability in terms of computational performance and to validate its optimality by means of psychophysical comparison experiments with existing methods.