54
J. Eur. Opt. Society-Rapid Publ. 21, 36( 2025)
Figure 1. Photograph of a sample illuminated from the back side with a semi-transparent laser-processed plasmonic metasurface on the front side. The colors exemplify two laser-induced gamuts observed in reflection and transmission modes on a single sample.
6 _ BR corresponds to a gamut with low lightness contrast( L * between 25 and 55) and presents a chromaticity value( C *) of up to 40, however only in the yellow domain( yellow hue h *).
8 _ BR shows a higher gamut volume than the two previous ones, has chromaticity values( C *) up to 40 and lightness values( L *) ranging from 20 to 80, but is really centered around yellow and presents no achromatic color with high lightness.
On the other hand, Figure 2d shows a synthetic gamut designed to produce images of poor quality( named Bad). This gamut was constructed by randomly selecting 50 primaries using NumPy’ s uniform random generator within specified ranges: lightness values( L *) between 40 and 60, and chromaticity values( a *, b *) ranging from �2 to20. This approach resulted in a non-centered gamut with low chromaticity( C *), allowing to verify whether the question about image quality is well understood by observers since this gamut should be ranked last for the majority of cases. Figure 2e illustrates the gamut of a CMY inkjet printer simulated using a Yule-Nielsen color prediction model calibrated with 36 printed color patches( named Inkjet). This gamut exhibits a wide range of colors, extending to high chromaticity( a *, b *) values in all directions, but lacks contrast with lightness( L *) values, none of which being below 40. Figure 2f illustrates a gamut consisting of colors extracted from a portrait image( named Portrait). A set of 100 primaries within the convex hull of the gamut was selected to create this color palette. This figure corresponds to a high-contrast, low-color gamut, with lightness( L *) ranging from 15 to 90, and chromaticity( a *, b *) going as high as 50, but only in the red / orange region, which is closest to skin tones.
Twenty-four different images were utilized for the subjective IQA based on questionnaires. Each image was reproduced using the six selected gamuts described above and sent through a survey for ranking by the observers on their personal display. In order to reduce the influence of the image content on the quality rating of the images to be assessed, our study included images representing various
Figure 2. Six gamuts used for the study presented in the CIE a * b * color plane with luminance color code.( a-c) correspond to laser-processed metasurfaces observed in back side reflection.( d) represents a synthetic small gamut.( e) represents a full color gamut of an inkjet printer without the black ink. Figure( f) represents a gamut which colors are extracted from a portrait image.
cases( see Fig. 3). Eight portrait images( Figs. 3a – 3d) and( Figs. 3i – 3l) were presented on a neutral background, similar to that found on identity documents. These images showed individuals with different skin tones. Among other images, some show a colorful content whose main colors can be inferred due to their spatial distribution( e. g. Fig. 3e). Some others have a colorful content whose main colors can hardly be inferred from their spatial distribution( e. g. Fig. 3f). Some images were selected because of their dull colors, or because their gamut is mainly located in one hue sector such as Figure 3w with only green colors. The image shown in Figure 3u is black and white and serves as a contrast and white achromaticity checker. The preliminary colorfulness evaluation of each image was performed using the Hasler colorfulness metric [ 34 ], and the resulting values are presented in Table 1.
The simulated images that were presented with the question format( Appendix A) to the observers for image quality evaluation were generated by employing a gamut mapping algorithm [ 1, 8 ] on the original image, as described in the introduction. This algorithm maps the colors present