利用影像统计信息选择适用的色域对映模式(英文)

Content-based Image Analysis for Automatic Gamut Mapping Selection

色域对映是跨媒体色彩复制的关键处理流程之一,过去的研究显示色域对映算法的表现与原稿内容有密切的关系。前人企图发展一套能够满足不同影像需求的算法。然而对于内容相异的图像而言,这些复杂的算法仍各有长短。因此,提出一套根据影像内容分析,选择最适算法的机制,希望藉由影像最适算法的选择,提升色彩复制系统的整体表现。为了在原稿输入时作实时的影像分析,必须设法降低影像统计的参数数量。为此,提出一套萃取影像关键统计因素的方法。在影像的统计上,撷取愈小的面积,运算成本愈低。研究结果显示,只要在影像的一些重要区域上作关键参数的统计分析,就能够以较低的运算成本,准确预测各影像的最适色域对映算法。

Gamut mapping is one of the key elements for cross-media color reproduction. Previous studies showed that the optimal process is almost invariably depended on the characteristics of the originals. Many attempts have been made on deriving a single gamut mapping algorithm (GMA) to adapt to a variety of images. However, the performance of the GMAs is still not steadfast. Therefore, a multi-GMAs solution which automatically selects the best-suited GMA from a range of them based on image analysis is proposed. To perform the analysis in real-time, one has to reduce the number of variables for image statistics. To this end, the approaches to extract limited image statistics to maximize the accuracy on predicting the best-suited GMA for given images are introduced. The results suggest: using a proper scanning pattern with the principal factors for image statistics, the quality of gamut mapping can be improved to a certain level.