Dental shade matching assisted by computer vision techniques

ABSTRACT

Accurately defining the colour of human teeth is important for restorations and dental treatments aimed at restoring the natural tooth colour. Visual techniques aided by digital cameras can be used to perform such task; however, this method still prone to errors due to the human eye subjectivity. Alternatively, a computer vision system can be used to automatically perform the colour matching process in digital photography. For such, it is fundamental to have high colour accuracy in all processed images. This paper presents two complementary techniques: (i) one that uses a camera characterisation algorithm that allows the device to generate photos with colorimetric responses and (ii) another that utilises the computed CIE L*a*b values to perform the matching process over shade guide colour samples to find the best match. Our results showed high agreement in colour matching after applying the device characterisation step with a [3undefined] polynomial, an average $\mathrm{\Delta }E$ below 1.64 units was obtained. Our matching method was tested on a labelled database that was created with photographs captured and annotated by an experienced specialist utilising VITA Classical Shade Guide. Seventeen photos were captured with a Canon EOS REBEL T5, whereas 20 photos were acquired with a Canon EOS REBEL T2i with cross-polarised lens, all utilising RAW mode. After processing and classification stages, weighted kappa was 0.74 and 1.00 for the Canon EOS REBEL T5 and Canon EOS REBEL T2i cameras, respectively. The performance of the proposed system with cross-polarised lens showed technical potential to be used as a tool for dentists to identify tooth shades accurately.

KEYWORDS:

• Colour matching
• colour selection
• colour calibration
• computer vision
• shade guide
• camera characterisation

Acknowledgments

The authors are thankful to the National Council for Scientific and Technological Development (CNPq), grants #132996/2020-0 and #309330/2018-1, for their financial support.

Disclosure statement

No potential conflict of interest was reported by the author(s).