利用色導表輔助相機測色的精確度評估

Performance of Camera-based Color Measurement using a Reference Color Chart

本研究之目的在於探討以下問題：（1） 手機相機測色直接按色域規格轉換的準確度如何?（2） 使用色導表輔助測色的準確度是否能提高?（3） 使用不同的迴歸模式校正色彩，是否能提高測色準確度?（4） 不同的相機、光源、樣本來源、擺放方式，對測色準確度的影響有多高?。研究結果顯示直接按色域規格轉換的誤差相當大，平均超過10ΔE00，使用色導表輔助測色可使色差降低一倍左右。色彩校正方法以「色差加權高階線性迴歸」最出色。塑膠色樣本由於具有半透明特性，因此需要在XYZ 色空間做「漏光增益補償」才能獲得較低的預測色差。在這相機、光源、樣本來源、擺放方式的表現方面，相機的表現各有千秋，四種受測光源下，D65 與CWF 表現較佳，TL84 容易凸顯同色異譜問題，表現最差。三種樣本中，雪面噴墨色差較小，塑膠色差較大。對塑膠而言，平躺的擺放方式比貼背要好。

The purpose of this study is to answer the followingquestions: (1) How accurate is the color measurementof mobile phone cameras by standard color spaceconversion? (2) Can the accuracy of color measurementbe improved by using a reference color chart? (3) Usingdifferent regression modes to correct color, can colormeasurement accuracy be improved? (4) How muchinfluence do different cameras, light sources, sampletypes, and placement methods have on the accuracyof color measurement? The experimental results showthat the error of direct color space conversion is quitelarge, with an average of more than 10 Δ E00. Using thecolor chart as a reference to assist color measurementcan reduce the color difference to a half. The bestcolor correction method is 'color difference-weightedhigh-order linear regression'. Due to the translucentcharacteristics of plastic color samples, it is necessary toperform a compensation for 'light leakage' in the XYZcolor space to reduce the color differences. In terms ofthe effects of the camera, light source, sample type, andplacement method, the results show that the performanceof the cameras varies. Under the four tested light sources,D65 and CWF perform better, while TL84 easilyinvokes the problem of metamerism and performs theworst. Among the three type of samples, colors of inkjetprinting on matte paper are most predictable and stable,whereas colors of plastic samples are most unpredictable.For placing the plastic samples, laying flat is better thansticking to the back in the viewing booth.