1

Introduction

Digital imaging for use in color matching of craniofacial and dental materials is a topic of present interest and continual development. Digital cameras are user friendly, relatively inexpensive, and readily available. The color information obtained from digital images can be easily manipulated into information that is relevant in a dental setting. For these reasons digital cameras are currently a good target device for the advancement of color matching in dental and craniofacial settings. Information obtained from a digital camera is generally input in a RGB color space that is device dependent and must be converted to a device independent standard color space . Because this information is device dependent, accurate color data management requires appropriate adjustment and calibration in order to be able to accurately utilize the color information extracted from a digital image . Because there are so many digital imaging devices that can be used, obtaining accurate color information that is device independent is necessary in order to standardize color information.

There are several categories of devices that are currently used in clinical settings for shade matching in dentistry. Spectrophotometers are considered to be amongst the most accurate and functional devices for these purposes . These devices measure the amount of light energy that is reflected from a specific object along the spectrum of visible light in 1–25 nm intervals . A spectrophotometer is made up of an optical radiation source, a light dispersing source, an optical measuring system, a detector, and a means of converting light to a signal for analysis and manipulation that is useful to the investigator . In the dental setting, the measurements obtained clinically are often compared to shade guide data to find the optimal shade tab counterpart . Colorimeters are another device used for color measurement and shade matching. The data that they acquire is often very precise because contact is made with the actual tooth . However, these devices do not measure spectral reflectance and are less accurate than spectrophotometers . Conventional colorimeters that utilize CIE recommended geometries for reflection measurements are generally not best for use of measuring objects with a translucent nature due to inaccuracies caused by the optical phenomenon of edge effects .

For color analysis involving digital camera sources, conversion equations from the red, green, blue (RGB) color space system to the CIELAB color space system are useful and necessary . Once the RGB values are converted to XYZ values, they can be easily converted to the CIELAB values .

Because RGB data obtained from a digital camera is device dependent, there is a need for the design and constant refinement of more complex calibration models that render optimal accuracy . In a study by Wee et al., four calibration models were evaluated with three different digital cameras using the cameras’ RGB values compared to the CIELAB values as a reference standard for accuracy measurements defined by ΔE . A second order polynomial regression (PRM2), a second order polynomial regression with eleven terms (PRM2-11), a third order polynomial regression (PRM3), and a model based on tetrahedral interpolation (TI) technique were all compared for accuracy .

In general, accuracy was improved by increasing the terms and raising the order of the regression model with proper terms being more important than increase of terms . In this study, three out of 12 calibration/camera pairs were found to be below the ΔE acceptability limit of 2.1 lending to the idea that inexpensive digital cameras used in combination with specific calibration methods have potential in the clinical processes involving color replication .

The specific objectives of this study were to develop a regression model that would relate digital RGB data from the VITA Linearguide 3D Master and VITA Bleached Guide 3D Master shade guides with their spectroradiometric correlates and to define a methodology of validating a digital image system for color accuracy. An additional aim was to provide summary RGB data and its relationship to lightness for the VITA Linearguide 3D Master and the VITA Bleached Guide 3D Master shade guides.

2

Materials and methods

Each of thirty five shade tabs included in a VITA Linearguide 3D Master shade guide and each of fifteen shade tabs included in a VITA Bleached Guide 3D Master shade guide were characterized by measuring the radiant energy using the spectroradiometer (PR 705; Photo Research Inc, Chatsworth, Calif) and a Xenon arc lamp (300W; Oriel Instruments, Stratford, Conn), with a connected fiber optic light cable. The spectroradiometer and optic light cable were fixed onto an optical table (Mecom Inc., Rising Sun, Ohio) and were placed inferiorly to the horizontal plane at a 45° angle in order to result in a 0° observation and a 45° illumination optical configuration for measurement. A measurement of radiant energy was produced for this central point on the standard from 380 nm to 780 nm in increments of 2 nm (Spectrawin 2.0; Photo Research Inc.) . The central ninth of each shade guide tab was measured. The radiant energy data was immediately converted to spectral reflectance and then to CIE XYZ values .

Digital images were taken of the same points on the shade tab specimens using a Canon Rebel T3i with an aperture setting of F22, an ISO of 200, a shutter speed of 1/200, the focal length set to 0.39, and the white balance set to a standard gray card. A Canon Ring Lite MR-14EX flash was used with a ¼ flash output. RGB values were extracted from the images of the central ninth of each of fifty shade tabs with Image J Image Analysis Software. Two to three measurements of RGB data from the images for each of eighteen shade tabs were taken using the central ninth of each shade tab. Duplicated shades within the two shade guides were treated as individual shade tabs and measured separately. Averages of multiple measurements were taken for each shade tab in its relative holder where multiple measurements were taken. Intraclass correlation coefficients for shade tabs that were measured multiple times and for shade tabs that were duplicated within the shade guides were determined using the SAS GLM procedure (PROC GLM, SAS ^® Proprietary Software 9.3, SAS Institute Inc., Cary, NC, USA). This procedure was done in order to determine the reliability of the RGB determination methodology used and in order to analyze the variability in same shade designations within the VITA Linearguide 3D Master and the VITA Bleached Guide 3D Master shade guides . The means of the RGB values were determined for each shade represented in the two shade guides used. These means were then used for the regression model determination. A linear regression model without y-intercept relating RGB values to XYZ values was generated using linear regression without intercept programming (PROC REG, SAS ^® Proprietary Software 9.3, SAS Institute Inc., Cary, NC, USA.) This regression utilized all of the thirty six distinct shades that were included in both shade guides. The R value was determined in order to indicate a level of correlation between the two systems of color description. Using this regression model, the RGB values were then converted to Commission Internationale d’Eclairage XYZ tristimulus values with D65 illumination and a 2° observer . The spectral reflectance data from the same specimens was also converted to XYZ values in the CIE standard color space . This regression would allow digital images to provide color information for a wide range of tooth color shades that can be used to accurately translate color information.

SAS 9.3 was used to determine summary RGB data and its relationship to lightness for each shade tab in the VITA Linearguide 3D Master and the VITA Bleached Guide 3D Master shade guides.

2

Materials and methods

Each of thirty five shade tabs included in a VITA Linearguide 3D Master shade guide and each of fifteen shade tabs included in a VITA Bleached Guide 3D Master shade guide were characterized by measuring the radiant energy using the spectroradiometer (PR 705; Photo Research Inc, Chatsworth, Calif) and a Xenon arc lamp (300W; Oriel Instruments, Stratford, Conn), with a connected fiber optic light cable. The spectroradiometer and optic light cable were fixed onto an optical table (Mecom Inc., Rising Sun, Ohio) and were placed inferiorly to the horizontal plane at a 45° angle in order to result in a 0° observation and a 45° illumination optical configuration for measurement. A measurement of radiant energy was produced for this central point on the standard from 380 nm to 780 nm in increments of 2 nm (Spectrawin 2.0; Photo Research Inc.) . The central ninth of each shade guide tab was measured. The radiant energy data was immediately converted to spectral reflectance and then to CIE XYZ values .

Digital images were taken of the same points on the shade tab specimens using a Canon Rebel T3i with an aperture setting of F22, an ISO of 200, a shutter speed of 1/200, the focal length set to 0.39, and the white balance set to a standard gray card. A Canon Ring Lite MR-14EX flash was used with a ¼ flash output. RGB values were extracted from the images of the central ninth of each of fifty shade tabs with Image J Image Analysis Software. Two to three measurements of RGB data from the images for each of eighteen shade tabs were taken using the central ninth of each shade tab. Duplicated shades within the two shade guides were treated as individual shade tabs and measured separately. Averages of multiple measurements were taken for each shade tab in its relative holder where multiple measurements were taken. Intraclass correlation coefficients for shade tabs that were measured multiple times and for shade tabs that were duplicated within the shade guides were determined using the SAS GLM procedure (PROC GLM, SAS ^® Proprietary Software 9.3, SAS Institute Inc., Cary, NC, USA). This procedure was done in order to determine the reliability of the RGB determination methodology used and in order to analyze the variability in same shade designations within the VITA Linearguide 3D Master and the VITA Bleached Guide 3D Master shade guides . The means of the RGB values were determined for each shade represented in the two shade guides used. These means were then used for the regression model determination. A linear regression model without y-intercept relating RGB values to XYZ values was generated using linear regression without intercept programming (PROC REG, SAS ^® Proprietary Software 9.3, SAS Institute Inc., Cary, NC, USA.) This regression utilized all of the thirty six distinct shades that were included in both shade guides. The R value was determined in order to indicate a level of correlation between the two systems of color description. Using this regression model, the RGB values were then converted to Commission Internationale d’Eclairage XYZ tristimulus values with D65 illumination and a 2° observer . The spectral reflectance data from the same specimens was also converted to XYZ values in the CIE standard color space . This regression would allow digital images to provide color information for a wide range of tooth color shades that can be used to accurately translate color information.

SAS 9.3 was used to determine summary RGB data and its relationship to lightness for each shade tab in the VITA Linearguide 3D Master and the VITA Bleached Guide 3D Master shade guides.