Estimation of spectral reflectance of object surfaces with the consideration of perceptual color space

Abstract

Due to the nonlinear transform from spectral reflectance to CIELAB values, the solution obtained by minimizing reflectance error in traditional spectral characterization methods will not be optimal when evaluated by colorimetric error. We propose a reflectance estimation method with consideration of perceptual color space. It combines two approaches, i.e., colorimetric-based spectral calculation and weighted spectral calculation. The experimental results show that the proposed method performs better than previous methods in terms of color difference with very slight degradation in spectral accuracy. © 2006 Optical Society of America OCIS codes: 330.1710, 330.1730 In many applications such as industrial color quality control and art painting archiving, there is an increasing need to estimate the spectral reflectance of object surfaces using a digital camera or a color scanner. Shi and Healey 1 presented a spectral characterization method to recover reflectance from scanner responses based on the linear reflectance model (LRM) and showed that it outperformed the traditional polynomial transform.2 Dicarlo and Wandell 3 found that the statistical distribution of spectral reflectance had an important influence on estimation accuracy. Inspired by that finding, Shen and Xin recently proposed two methods, namely, adaptive estimation 4 (AE) and optimized adaptive estimation (OAE), 5 for reflectance recovery by training sample selection and weighting. However, due to the color matching functions and the nonlinear transform between CIEXYZ values and CIELAB values, the minimization of reflectance error of previous methods does not guarantee an optimal solution in terms of color difference. This may be one of the reasons why the color accuracy of colorimetric characterization is sometimes better than that of spectral characterization. 6In this Letter, we propose a method for the accurate estimation of spectral reflectance from the responses of a color scanner with the consideration of approximately perceptually uniform color space CIELAB. To reduce errors in reflectance estimation, two approaches, namely, colorimetric-based spectral calculation and weighted spectral calculation, are involved in this method.We suppose that the visible spectrum, from 400 to 700 nm, is equally sampled in N wavelengths. For an ideal three-channel color scanner, its 3 ϫ 1 response vector u is simply the product of the 3 ϫ N matrix M s of spectral responsivity (including spectral radiance of the illuminant and spectral sensitivity of sensors) and the N ϫ 1 vector r of spectral reflectance, i.e., u = M s r. For a common scanner however, its 3 ϫ 1 actual response vector may carry some influence from a nonlinear optoelectronic conversion function (OECF) F OECF ͑·͒ and a 3 ϫ 1 response bias vector f caused by dark current. Hence the imaging process is formulated aswhere v = M s r + f. As F OECF ͑·͒ can be obtained using the actual responses and mean reflectance (or luminance) of gray-scale patches, 4 the crucial equation reads ͑2͒from which M s ...

Practical Radiometric Compensation for Projection Display on Textured Surfaces using a Multidimensional Model

“…It is impossible to recover r without any prior knowledge of the illumination and imaging system or any known targets [SX07]. Fortunately, many prior works have shown that spectral reflectance of most real‐world materials can be well represented by low dimensional data (usually 3 – 9) [PHJ89] (e.g.…”

Colour matching comparison between spectrophotometric and multispectral imaging measurements

“…A spectrophotometer is incapable of colour matching multicolour fabrics and samples of small sizes. With the development of digital imaging technology, multispectral imaging (MSI) systems have become available for extracting colour information of fabrics. An MSI system can provide not only the spectral information but also the spatial information of a sample compared with spectrophotometers.…”

An efficient method for solid‐colour and multicolour region segmentation in real yarn‐dyed fabric images

“…This process is inefficient and error prone owing to inconsistency in preparing yarn samples [1]. With the development of digital imaging technology, multispectral imaging (MSI) systems [3][4][5][6][7] are adopted to extract the colour information of a fabric sample. MSI systems can provide not only the spectral information but also the spatial information of a sample.…”