Correlation of Visual and Instrumental Color Measurements to Establish Color Tolerance Using Regression Analysis
视觉和仪器颜色测量的相关性，以使用回归分析确立颜色公差
简介
Color is one of the important parameter considered in the determination of quality for fashion materials like leather. The color variation in each piece of leather sample in a batch should be within the acceptable range. Visual assessment is currently used in leather industry for quality control and color sorting. The current method used is subjective and often leads to disagreement between buyer and seller. Color measurement using reflectance spectrophotometer evaluates color consistently and is an objective assessment system. However, there is always an apprehension that the instrumental color assortment may not agree with the human perception of color difference. Hence, in the present investigation, an attempt was made to screen four color difference formulae, viz., CIELAB76, CMC (2:1), CIE94 and CIE2000 for their suitability in obtaining pass/fail decisions, which would be in conformity with that of the average human observer.
Regression analysis was performed to find a correlation between visual and instrumental color assessments and the results indicate that CMC(2:1) formula may be the most suited for the purpose. A detailed analysis of visual and instrumental color values revealed visual non-uniformity towards sensitivity to hue, chroma and lightness. This difference in sensitivity was also taken into consideration in instrumental color sorting and pass/fail tolerance was established that led to closer conformity between visual and instrumental methods.
颜色是纺织皮革等时尚材料在确定质量时考虑的重要参数之一。每批皮革样品的颜色变化应在可接受范围内。视觉评价目前在皮革工业中用于质量控制和颜色分类。目前使用的方法是主观的，常常导致买卖双方的分歧。用反射分光光度计测量颜色是一个客观的评价系统。然而，人们总是担心，仪器测量的颜色进行分批和品管可能不符合人类对色差的感知。因此，在本研究中，我们尝试筛选出四种色差公式，即CIELAB76、CMC（2:1）、CIE94和CIE2000，以确定它们在获得颜色品管决策方面的适用性，以期与普通人类观察者的结果相一致。
回归分析发现视觉和仪器颜色评估之间的相关性，结果表明CMC（2:1）公式可能是最适合的。对视觉和仪器颜色值的详细分析表明视觉对色调、彩度和明度的敏感性不均匀。在仪器颜色分批和品管容差中也考虑到了灵敏度的差异，从而使视觉方法和仪器方法更加一致。
Materials
A commercial batch of dyed suede garment leathers (220 leathers) in grayish black shade from a leather manufacturing in Chennai, India was chosen for the study. The buyer’s sample was used as standard for assessment of color variation. The color on the suede side of the standard and each sample in the experimental batch of leathers is given in Figure 1. 221 leather samples were separately compared for visual as well a instrumental color analysis.
材料
以印度钦奈市一家皮革厂生产的染色麂皮服装革（220张）为研究对象。买方的样品被用作评估颜色变化的标准。图1给出了标准革和试验批皮革中每个样品的绒面颜色。221个皮革样品分别进行了视觉和仪器颜色分析的比较。
Methods
Visual Assessment of Color Tolerance Twenty experienced color matching experts of leather industry were selected for the study and D65 light source representing average daylight was used for viewing. The standard was compared against each leather sample and visual color matching was decided as “pass” or “fail” as per the color perception of each expert. The assessment of leather samples made by the experts was recorded as ‘0’ for fail and ‘1’for pass. Logit regression analysis25 was used to convert the recorded binary data response in terms of visual difference (DV) using Equation 8, which is a modified form of the equation reported by H. Mangine et.al.
方法
色差视觉评价选用20名皮革行业有经验的配色专家进行研究，采用代表平均日光的D65光源进行观察。将标准与每一个皮革样品进行对比，根据每位专家的色觉判断视觉配色为“合格”或“不合格”。专家对皮革样品的评估记录为“0”表示不合格，“1”表示通过。使用Logit回归分析25将记录的二进制数据响应转换为视觉差异（DV），该公式是H.Mangine报告的方程的修正形式等.
Si is the number of experts who found the match acceptable and Ni is the total number of experts. The main difference between the current modified Equation 8 and the equation reported by Mangine et al. is the use of constant 2, which was obtained by trial and error method to get positive DV value as well as meaningful correlation with instrumental DE values.
Si是认为匹配可以接受的专家的数量，Ni是专家的总数。当前修正的方程8与Mangine等人报道的方程之间的主要差异。是利用常数2，通过试错法得到正的DV值以及与仪器DE值有意义的相关性。
In this study a detailed color analysis for grayish black color data with lightness value (L* ranging from 19 to 28) varying in hue and chroma was carried out. Pass/Fail decisions based on single number ΔE* ab value (ΔE >1 as fail and ΔE <1 as pass) does not correlate with visual assessment of color. Three dimensional tolerance limits based on L, C, h color space provides closer agreement to visual perception. The methodology for setting tolerance limits using two dimensional grouping based on hue and chroma along with lightness difference provides an objective solution for color sorting problem. Tolerance limits in terms of L,C, h vary for different spectral colors. Instrumental color sorting helps the buyer/seller to fix the tolerance of each color parameter (i.e. in terms of hue, chroma, lightness) providing consistent sorting which results in accurate and faster pass/fail decisions.
在这项研究中，对亮度值（L在19到28之间）在色调和色度上变化的灰黑色数据进行了详细的颜色分析。基于单个数字ΔEab值（ΔE>1表示失败，ΔE<1表示通过）的判定与颜色的视觉评估无关。基于L、C、h颜色空间的三维公差限制更接近视觉感知。基于色调、色度和明度差的二维分组方法，为颜色分类问题提供了一个客观的解决方案。L、C、h的公差限值因光谱颜色不同而不同。仪器颜色分类有助于买方/卖方确定每个颜色参数（即色调、色度、亮度）的公差，提供一致的分类，从而实现准确、快速的通过/失败判定。
Conclusions
Mathematical models were developed to correlate visual color difference and instrumental methods based on four color difference equations. Both t-test and coefficient of determination (R2 ) showed that the equations from best to worst fit follows the order CMC(2:1) > CIE94 > CIEDE2000 > CIE76. For a substrate like leather with non-uniform and porous surface, CMC equation may be the best suited to correlate with visual color perception. A detailed analysis of the color values established that, for an average expert the order of sensitivity is arranged as hue > chroma > lightness.
An objective solution for color sorting problem has been found by setting tolerance limits based on hue-chroma (two dimensional grouping) along with lightness led to closer conformity between visual and instrumental methods.
结论
基于四个色差方程，建立了视觉色差与仪器方法之间的数学模型。t检验和决定系数R2表明，方程组从最佳到最差的拟合顺序为CMC（2:1）>CIE94>CIEDE2000>CIE76。对于具有不均匀和多孔表面的皮革基材，CMC方程可能最适合与视觉色觉相关。对颜色值的详细分析表明，对于一般专家来说，敏感度的顺序是色调>色度>亮度。通过根据色调色度（二维分组）和亮度设置公差限值，使视觉和仪器方法更加一致，找到了一种客观解决颜色分类问题的方法。
