Modelling texture appearance of gonioapparent objects

Quantifying the appearance of coating products is essential in the automobile and
automobile finishing industries for efficient product development and product/quality
control. There is a specific need to develop techniques to measure the total appearance
of metallic coatings. The present study focuses on two key attributes of visual texture:
coarseness and glint. In order to develop models to capable of measuring these
attributes, it was first necessary to design psychophysical experiments for assessing
coarseness and glint as perceived on metallic-coating panels. The change in the
appearance of the metallic coatings is known as a gonioapparent effect, and is greatly
dependant on the illumination and viewing conditions. Therefore, appropriate
conditions were carefully examined for the independent observation of coarseness and
glint in order to discern those attributes. It was found that diffuse illumination was
appropriate for viewing coarseness and directional illumination was appropriate for
observing glint. Under these appropriately-controlled conditions, the perceptual
coarseness and glint of sets of metallic-coating panels were assessed by human
observers.
A digital camera was used to capture information on the spatial detail of the
metallic-coating panels. An image of each panel was captured under the same viewing
conditions as used for the visual assessments. The information in a single image was
sufficient to represent a metallic-coating panel under identical diffuse illumination
conditions for which observers assessed coarseness. For capturing information on glint,
however, a high dynamic-range (HDR) image was necessary because the dynamic range
of the scene in which the glint was observed exceeded that of the camera used in this
study. Two computational models were developed to predict perceptual coarseness and
perceptual glint by extracting associated features from the captured images. The
performance of these models was verified by comparing predictions made using them
with the perceptual coarseness and glint scaled by observers.
For industrial use, the visualisation of products on computer displays would give
various opportunities, for example, to develop and design products or coatings and also
to communicate appearance information. A digital camera and a suitable display would
enable this to be achieved, but the ability to reproduce the appearance of metallic-coating
products on displays in a satisfactory manner was found to have significant
~
challenges. The coarseness model developed in the present study was able to represent
perceptual coarseness based on the images captured by the digital camera. However,
the resolution of the images was not high enough to resolve the individual aluminium
flakes contained in the coatings, which contribute to the visual texture. Therefore,
verification of the images was carried out for the coarseness attribute by comparing the
coarseness perceived in the images displayed on a liquid-crystal display (LCD) with the
metallic-coating panels themselves. In addition to camera limitations, LCD resolution
also prevented the same conditions used for physical panel assessment from being
replicated. Therefore, two optimal conditions were selected and perceptual coarseness
was scaled using images. Besides the difference in experimental conditions, there was
also a difference in the "absolute" texture appearance between the two media because of
errors in image reproduction of the images. In spite of this, the relatively-scaled
perceptual coarseness for the image samples agreed well with that for the original
physical samples. This implies that it is practicable to assess perceptual coarseness
from an image on a display that simulates a metallic-coating panel