Ruili He has successfully defended her thesis!
We are very happy to announce that ApPEARS ESR Ruili He has successfully defended her thesis titled ‘Accurate Colour Reproduction of Human Face using 3D Printing Technology’ at the University of Leeds, May 22, 2023.
Abstract of thesis
The skin colour of the face is one of the most significant factors in appearance and perception of an individual’s health, age, attractiveness, emotional state, ethnic and cultural identity, etc. With the rapid development of colour 3D printing technology and 3D imaging acquisition techniques, it is possible to achieve skin colour reproduction with the application of colour management. However, due to the complicated skin structure with uneven and non-uniform surface, it is challenging to obtain accurate skin colour appearance and reproduce it faithfully using 3D colour printers.
The aim of this study was to improve the colour reproduction accuracy of the human face using 3D printing technology. A workflow of 3D colour image reproduction was developed, including 3D colour image acquisition, 3D model manipulation, colour management, colour 3D printing, postprocessing and colour reproduction evaluation. Most importantly, the colour characterisation methods for the 3D imaging system and the colour 3D printer were comprehensively investigated for achieving higher accuracy. Specifically, a designated digital imaging system was developed to capture human faces to explore the effects of different factors, such as image format, training dataset and mapping method, on the accuracy of skin tone estimation. The findings supported the successful colour characterisation for the 3dMDface system to capture a 3D model of the human face with realistic skin colours.
For the colour 3D printer, the methods of conventional polynomial regression and deep neural networks were investigated for printer colour characterisation which transforms from the printer-controlled colour space to an output device-independent colour space, and five types of variables were compared as the output for producing higher precision in printing skin colours. Moreover, a specific skin colour chart was printed using the 3D printer to determine the colour profile for achieving faithful skin colour reproduction.
In addition, psychophysical experiments with the grey-scale method were conducted to collect visual colour-differences data of 3D printed samples, and the results showed that the performance of CIELAB and CIEDE2000 colour-difference formulae needs to be improved for 3D objects. Consequently, the parametric factors in the colour-difference formulae were optimised ( for CIELAB, and for CIEDE2000), and the accuracy of 3D colour reproduction was evaluated using the optimised formulae.
Based on the developed colour characterisation methods, a 3D model of a human face was printed using the proposed 3D colour image reproduction workflow, and the accuracy achieved was 5.73 CIELAB units (3.88 CIEDE2000 units). The values were 3.75 and 3.42 units for the optimised colour-difference formulae. Overall, the methods and results presented in this thesis provide a new and improved understanding of colour characterisation and colour evaluation in 3D colour reproduction of human faces.
Ruili He at the graduation Ceremony at the University of Leeds