Abstract : This paper proposes a rendering framework that visualizes the phenomenon of color fading due to light exposure in a painterly style. Unlike conventional non-photorealistic rendering techniques that merely mimic the colors or textures of the original image, our approach reflects not only lighting-induced color shifts and material properties but also gradual color fading over time, thereby generating more realistic and vivid painterly imagery. The proposed method begins by approximating a 2D image into a 3D structure to estimate normal vectors and performs relighting based on shading values determined by user-defined lighting conditions. Then, using empirically measured pigment fading data, a tetrahedral color space is constructed and simplified through graph optimization for efficient color interpolation. Strokes are automatically generated based on color differences and image moment analysis to determine their position, size, orientation, and color. The final output is produced through brushstroke-based painterly rendering. Our method has been applied not only to still images but also to videos, hand-drawn illustrations, and neural style transfer outputs. We conducted comparative analyses of rendering results under varying lighting conditions, degrees of color fading, and graph simplification levels. Additionally, by combining our technique with Neural Style Transfer, we successfully integrated natural fading effects into existing stylized images. Experimental results demonstrate the superiority of our approach in terms of temporal coherence, color expressiveness, and visual plausibility. These findings suggest broad applicability in areas such as cultural heritage visualization, digital restoration, and interactive media art.

[paper]