This paper investigates the problem of appearance-based virtualization of physical objects in spatial augmented reality, where projected imagery should simultaneously preserve object recognizability, integrate artistic style, and maintain color harmony with a virtual background. We implemented two baseline methods—Average Color, Texture Overlay— and a transformer-based style transfer approach (StyTr \(^2\) ) and conducted a user study evaluating them across three perceptual criteria: contour clarity (Q1), painterly touch integration (Q2), and color harmony (Q3). The results show that while Average Color excels at contour preservation, and Texture Overlay is superior for stylistic and chromatic integration, only StyTr \(^2\) achieves a perceptual balance across all criteria. Quantitative analysis confirms that StyTr \(^2\) yields the most stable performance with the lowest overall variance. Notably, it received particularly high ratings for painterly touch in scenarios featuring fine-grained texture, such as pointillist styles, suggesting its effectiveness in reproducing style-specific surface qualities.