Development of Image-dependent Gamut Mapping Algorithm

本研究旨在开发一种跨不同颜色媒体的图像相关色域映射算法（GMA），重点关注以下五个主要项目。(A)基于图像相关概念的非线性色域映射模型的基本设计。(B)3D图像到设备色域映射算法（3D I-D GMA）的开发c针对窄色域图像的色域扩展算法的研究。(D)基于图像关键色提取的悦色变换算法研究。(E)图像的紧凑三维色域边界描述符（GBD）的开发及其在三维I-D GMA中的应用我们在2000年、2001年和2002年期间的研发结果分别总结在（1）图（3）、（4）图（6）和（7）图（9）中。(1)图像相关GMA和色调划分的非线性2DI-D GMA的概念模型（2）3D图像色域壳表面的自动形成（3）用Overhauser样条函数描述平滑的3D色域壳表面（4）用所提出的3DI-D GMA进行的映射实验（5）用r-图像方法压缩的3D色域边界描述符（GBD）（6）基于高斯直方图规范的色域扩展算法（7）（8）基于奇异值分解和小波图像压缩技术的三维r图像GBD压缩（9）基于r图像GBD的三维I-D GMA的心理物理实验总之，三维I-D GMA的性能优于传统的二维D-D GMA、裁剪GMA和典型的三维D-D GMA。所提出的I-D GMA算法对于高饱和度连续色调的宽色域CG图像效果非常好。此外，一种新的GBD的r图像方法是非常有用的图像和设备之间的色域大小的快速比较。

This research has been addressed to develop an Image-dependent Gamut Mapping Algorithm (GMA) across the different color media focusing on the following five major items.(A) Basic design of nonlinear gamut mapping model based on the concept of Image-dependence.(B) Development of 3D Image-to-Device Gamut Mapping Algorithm (3D I-D GMA)c Research on Gamut Expansion Algorithm for narrow gamut images.(D) Research on Pleasant Color Transform Algorithm by key color extraction from image.(E) Development of compact 3D Gamut Boundary Descriptor (GBD) for image and its application to 3D I-D GMAOur R&D results during 2000, 2001, and 2002 are summarized in (1)〜(3), (4)〜(6) and (7)〜(9), respectively as follows.(1) Conceptual model of Image-dependent GMA and Hue-divided nonlinear 2D I-D GMA(2) Automatic formation of 3D Image Gamut Shell Surface(3) Description of smoothed 3D Gamut Shell Surface by Overhauser Spline Function(4) Mapping Experiments by the proposed 3D I-D GMA(5) Compact 3D Gamut Boundary Descriptor (GBD) by r-image method(6) Gamut expansion algorithm based on Gaussian Histogram Specification(7) Pleasant color conversion by Principal Component Matching in segmented colored areas(8) Compression of 3D r-image GBD by SVD and Wavelet image compression techniques(9) Psychophysical experiments on 3D I-D GMA using r-image GBDIn conclusion, 3D I-D GMA resulted in the better performance than the conventional 2D D-D GMA, Clipping GMA, and typical 3D D-D GMAs. The proposed I-D GMA worked very well especially for the wide-gamut CG image with highly saturated continuous tone. Besides, a novel GBD by r-image method was very useful for the quick comparison of the gamut sizes between image and device.