AIC Lisboa 2018

Interim Meeting of the AIC:

crow logo2

Sept 25-29, 2018. Lisbon, Portugal

The aim of this Interim Meeting is to bring together a multidisciplinary group of specialists who use colour and light in practice or/and theory within a broad range of scientific, technological and artistic domains with a focus on human comfort.

Full meeting details here: http://www.aic2018.org/

Contributions from CRSC Members:

Colorization of Dichromatic Images
Brian Funt1 and Ligeng Zhu
School of Computing Science, Simon Fraser University

1CRSC Board Member. Website: http://www.cs.sfu.ca/~funt/

Abstract: What it is like to be color blind? What colors do dichromats see?  How much 'color' is really missing from a dichromat's experience. Since it has been reported that many people do not realize they are color blind until they are many years old, perhaps the difference is not so significant. Of course, we likely can never know what another person experiences, but we can explore what color information dichromatic vision provides. In recent years, many colorization methods have been described in the computer vision literature.  Given only a greyscale (i.e., luminance) image, these computer-based colorization methods generate a color image with very believable colors. Colorization methods are generally based on 'deep learning' the connection between luminance, the context, and probable color. It appears they encode knowledge about the world such as clear sky is blue, clouds are grey, beaches are a sandy color, forests and grass are green, and so forth.  Since colorization works for luminance images, we explore how well they might work for dichromatic images.  The results of colorizing dichromatic images can be expected to give us more insight into what color information is present, as well as missing, for the dichromat at a more experiential level than the standard statements that deuteranopes -- observers lacking M cones--- cannot distinguish some reds and greens from one another.  We employ a modification of the colorization method of the Iizuka et al. ("Let there be color!: Joint end-to-end learning of global and local image priors for automatic image colorization with simultaneous classification," Proc. of SIGGRAPH 2016, 35(4):110:1-110:11, 2016) to colorize dichromatic images. In particular, given an image in sRGB (non-linear), it is converted to linear XYZ (CIE XYZ space), and then to LMS cone space using the Hunter-Pointer-Estevez transformation matrix. To simulate a deuteranope, the M channel is discarded, yielding an LS image. The neural network is trained on a set of 50,000 images from the Microsoft COCO image dataset. The training involves inputting the LS image and the corresponding LMS image. During training, the network's weights are adjusted so that M is predicted from the L and S. The training converges to a low average error. The network's LMS output is then converted back to sRGB for display. Visually, the results are surprisingly believable. In other words, one does not easily recognize that a colorized dichromatic image is, in fact, 'dichromatic' (i.e., in the sense that it is derived from only two channels of color information), and not a regular full-color image.

Color Discrimination Ellipses Explained by Metamer Mismatching
Brian Funt1 and Emitis Roshan
School of Computing Science, Simon Fraser University

1CRSC Board Member. Website: http://www.cs.sfu.ca/~funt/

Abstract: Many psychophysical experiments have shown that color discrimination varies substantially with the region of color space in which the colors reside. Many models of the experimental data have been proposed, and many uniform color spaces have been developed that attempt to represent color in a coordinate system such that equally discriminable colors are equal distances apart.  All of these models and uniform color spaces are based on fits to the experimental data.  Many provide good fits to the data, but they remain data models and do not explain why color discrimination varies in the way it does.  In contrast, we propose a theory of color discrimination based on metamer mismatching. Metamer mismatching refers to the extent to which two physically distinct reflectances that match under one light fail to match under a second light.  The study by Zhang et al. (Zhang, X., Funt, B. and Mirzaei, H., "Metamer Mismatching in Practice versus Theory," Journal of the Optical Society of America A, Vol. 33, No. 3, pp. A238-A247, March 2016.) of metamer mismatching showed that it is most severe for grey and least severe for highly saturated colors. Our hypothesis is that in order to be able to reliably discriminate physically distinct surfaces from one another observers must be more sensitive to the differences between colors for which metamer mismatching creates significant uncertainty (i.e., when the metamer mismatch bodies are large), and least sensitive for colors for which metamer mismatch creates little uncertainty. In particular, we propose that the sensitivity of color discrimination is inversely related to the degree of metamer mismatching. It is common to represent color discrimination in terms of ellipsoids in color space and ellipses in chromaticity space. One relatively recent set of experimental data on color discrimination is that of Huang et al. (Min Huang, Haoxue Liu, Guihua Cui and M. Ronnier Luo, "Testing Uniform Colour Spaces and Colour-Difference Formulae Using Printed Samples," Color Research & Application 37(5), October 2012). For discrimination ellipsoids of volume Vd measured by Huang et al. and metamer mismatch bodies of volume Vm, we find that Vd is linearly proportional to 1/Vm with a correlation coefficient of 0.76. This clearly supports the hypothesis that the need to overcome the uncertainty due to metamer mismatching is the reason for more precise discrimination between colors in some regions of color space. Since Zhang et al. showed that metamer mismatching is greatest for grey, high for colors of low saturation, and decreases with increasing saturation, our hypothesis correctly predicts that color discrimination is finest near grey and coarsest for the saturated colors near the object color solid boundary. In other words, metamer mismatching provides an explanation for why color discrimination varies in the way it does.

Visual Experience of Older Adults: Colour Preference and Colour Tuning for Health and Comfort

Sharyn, Adler Gitalisa,b

aOCAD University, Toronto, Canada; bCRSC Board Member

This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract: Understanding visual conditions of the aging population helps us to define major challenges faced in navigating space and living healthy lives. For older adults colour and light can be used strategically to improve quality of life. The global population aged 60 or over is growing faster than all younger age groups. According to data from World Population Prospects: the 2017 Revision, (U.N.) the number of older persons -those aged 60 years or over- is expected to more than double by 2050 and to more than triple by 2100. One of the five strategic priority areas recognized by the World Health Organization, Global strategy and action plan on aging and health 2016-2020, is Age Friendly Environments. As the population ages rapidly, we are faced with the opportunity to create living environments that promote independent healthy lives.

Biological and environmental factors influence the older adults’ spatial orientation, object recognition and social behavior. Incorporation of daylight and electric light into the built environment can help compensate for the diminished visual system and lifestyle changes of the older adult.

Effective use of lighting can lessen the impact of the aging eye on task visibility; changes in balance and postural stability leading to higher risks of falls; adjust hormonal balance, behavior and mood; improve changes in the circadian system that can impact the quality of sleep. Quality lighting design that is strategically incorporated with colours and materials in interior design enhance visual ergonomics and wayfinding. Intentional use of CCT and quantity of light helps older adults experience colouration designed intentionally to create comfortable living spaces. Quantity of light and colour fidelity boosts appetite and mood. Quantity of light and spectral sensitivity is critical for the regulation of the circadian system.

Changes and developments in lighting technology present unprecedented opportunities, but there are also risks and challenges for the new technologies. One of the biggest trends in lighting is the development of colour tuning LED fixtures. Individuals have the ability to decide with one light source the quantity of light and what Kelvin temperature of light they want to experience ranging from candlelight at 1700K to daylight at 6500K. Is incorporating tunable white LED fixtures in people’s homes the advisable route for aging adults? If a spectrum of the light is making colours look less attractive to them, should the source change to make colour more vivid and interesting? Or conversely, is the best route to change the colour in a space, for example using the aging person’s preferred colours in paint colour, objects and fabric of a space? This talk will examine how aging people see colour, colour preference of aging adults and discuss the benefits and risks of colour tuning light.

 

Print Email


Send Email

aic color
Canadian member of the International Colour Association