逆光人像印刷階調複製偏好之研究

A Study on Tonal Reproduction Preferences in Backlit Portrait Printing

在逆光環境下拍攝的人像若印製於色域較窄的非塗佈紙上，可能導致背景與人臉輪廓模糊不清，影響視覺體驗。由於影像辨識技術日漸成熟，我們可以運用人工智慧來判定待複製影像是否屬於逆光人像，並利用影像分割技術，對背景與前景分別進行不同的階調調整，使影像即便在動態範圍受限的印刷環境中，仍能呈現豐富的紋理細節。本研究透過人因實驗，探討逆光人像在印刷複製過程中的階調偏好，並提出適用於逆光人像印刷的RGB影像前處理模式，以提升最終成品的視覺效果。本研究模擬不同逆光人像在非塗佈紙列印環境下，調變背景與前景γ值的階調調整效果，人因視覺偏好度的實驗結果顯示，平均而言，前景/背景γ值0.8/1.2的組合表現最好。此外，前景/背景的明度值可用預測理想的γ調整值。然而，由於影響因素眾多，預測的精確度仍有待改進。

This study aims to explore the tonal preferences of backlit portraits in the printing reproduction process and propose an RGB image preprocessing method suitable for backlit portrait printing to enhance the final product's visual quality. With the advancement of image recognition technology, artificial intelligence can be used to determine whether an image to be reproduced is a backlit portrait. Additionally, image segmentation techniques can be applied to separately adjust the tonal characteristics of the background and foreground, ensuring that the image retains rich texture details even in a printing environment with limited dynamic range. This study simulates the effects of tonal adjustments by varying the gamma values of the background and foreground in a backlit portrait printed on uncoated paper. The experimental results on human visual preference indicate that, on average, the combination of foreground/background gamma values at 0.8/1.2 yields the best performance. Furthermore, the mean lightness values of the foreground and background can be used to predict the ideal gamma values for image adjustment. However, due to the multiple influencing factors involved, the accuracy of such predictions still requires further improvement.