Virtual Color Restoration for Ancient Architecture by Color Transfer

Abstract

Virtual color restoration is an important work to protect or exhibit decorative paintings on ancient architecture. This paper proposes a novel method of colorful pattern segmentation and color restoration for ancient architecture based on improve color transfer algorithm. First, a color image and a shape image are converted from the RGB to $$l\alpha \beta$$ color space. Second, the color and shape images are segmented using the improved clustering algorithm of density-based GMM (Gauss mixture model). Third, the relations between the color and shape image regions are determined using the nearest region-matching algorithm to avoid the problem in which multiple areas are matched to the same region. Finally, color is transferred to the corresponding region in the shape image, and a merged image is obtained. Contrastive experimental results demonstrate that the proposed method can restore paintings of ancient architecture effectively more accurately and efficiently.