A Study on Bi-Directional Gamut Mapping with Gamut Compression or Expansion for HDR Image

我们一直在开发一种双向通用色域映射算法，通过从宽到窄的色域压缩或从窄到宽的色域扩展来有效地适应每个设备的色域，从而获得令人愉快的图像。以下是我们研究结果的摘要。我们用r-image表示图像和喷墨打印机、lbp等设备的色域外壳，并进行如下分析：(a)用r-image进行数值色域比较，(b)用亮度分割进行外色域特征分析。然后我们发现了色域映射的以下一般特征。(1)当一幅图像的r-image像素值在整个区域和中亮度区域附近的最大值大于设备的像素值时，需要进行色域压缩。(2)如果它们小于中间区域的设备，并且主要位于较低的亮度区域，则需要扩展色域。最后，构建了双向色域映射的基本算法。为了精确地描述少量彩色芯片的色域表面，我们开发了一种不产生空段的高精度r-image的映射算法，其思路如下：(1)对每个颜色空间进行不均匀分割，以包含恒定的颜色样本；(2)将色域表面塑造为多边形网格；(3)将多边形网格再次分割为恒定的离散极角段（Δθ，Δψ）。色域压缩/扩展算法的开发我们开发了一种图像相关的色域扩展算法，称为“直方图拉伸法”，并对典型图像样本进行了测试。指出了在亮度轴上确定扩展断点的重要性，并根据器件的色域边界和图像中颜色分布的标准差进行了统计确定。进一步，该方法发展为“直方图重标度法”，无需任何色域比较程序即可进行连续双向映射。通过对实际图像的实验，验证了自适应色域扩展和压缩可以在输入图像和设备之间自动进行。HDR图像的动态范围压缩是显示HDR图像的必要条件。我们使用集成的环绕场开发了改进的Retinex模型，并通过将32位HDR图像压缩为8位LDR图像，极大地改善了图像中暗处的颜色外观。此外，我们开发了一种LCRT（Local Contrast Range Transform）算法，并将LCRT应用于摄像机拍摄的HDR图像，成功地显示出高质量的图像。基于场景参考的愉悦图像色域映射算法的开发通过提出的基于直方图拉伸/缩放的色域映射，最终将色彩再现过程委托给CMS。另一方面，我们开发了“不同场景之间的颜色传输/颜色交换算法”作为一个新概念，以产生令人愉悦的彩色图像。我们发现，在没有任何常规彩色芯片和任何标准图像的情况下，通过不同场景之间的色域映射可以获得令人愉快的图像。少

We have been developing a bi-directional versatile gamut mapping algorithm with gamut compression from wide to narrow or gamut expansion from narrow to wide for obtaining a pleasant image by adapting to each device gamut effectively. The following is a summary of our results.1.Construction of bi-directional gamut mapping model and quantitative gamut comparison methodWe represented gamut shells of an image and devices such as inkjet printers and LBPs by r-image, and performed the following analysis : (a)numerical gamut comparison by r-image and (b)feature analysis of outside gamut by lightness segmentation. Then we found the following general characteristics of the gamut mapping. (1)When the most of r-image pixel values for an image are greater than those for device in entire region and around middle lightness region, a gamut compression is necessary. (2)If they are smaller than those for device in middle region and mainly located at lower region of lightness, a gamut expansion is desir … More able. Finally, we constructed a basic algorithm for bi-directional gamut mapping.2.Development of a high accurate description method of a device gamutIn order to describe the gamut surface with a small number of color chips precisely, we developed a mapping algorithm to highly precise r-image which did not produce any empty segments by the following ideas : (1)color space was non-uniformly divided into a segment for each to include the constant color samples, (2)the gamut surface was shaped as polygon meshes, (3)the polygon meshes were divided again by a constant discrete polar angle segment (Δθ,Δψ).3.Development of a gamut compression/expansion algorithmWe developed an image-dependent gamut expansion algorithm so called "histogram stretch method" and tested for typical image samples. We pointed out that decision of an expansion breaking point in a lightness axis was important, and decided the point statistically based on the gamut boundary of a device and the standard deviation of color distribution in an image. Furthermore, the method developed into "histogram rescaling method" which could perform continuous bi-directional mapping without any gamut comparison procedures. We verified that the adaptive gamut expansion and compression could be performed between an input image and a devise automatically through experiments with actual images.4.Examination about an application to an HDR imageDynamic range compression is indispensability to display an HDR image. We developed an improved Retinex model using an integrated surround field, and largely improved the color appearance at the dark place in the image by compressing a 32 bit HDR image into an 8 bit LDR image. Moreover, we developed an LCRT(Local Contrast Range Transform) algorithm and succeeded to display high quality images by applying the LCRT to HDR images captured by a video camera.5.Development of gamut mapping algorithm to a pleasant image by scene referenceBy the proposed gamut mapping based on the histogram stretch/rescaling, color reproduction process is finally entrusted to CMS. In another point of view, we developed "color transmission/color exchange algorithm between different scenes" as a new concept to produce a pleasant color image. We found possibility to get a pleasant image by gamut mapping between different scenes without any conventional color chips and any standard images. Less