Spectral Filter Selection for Increasing Chromatic Diversity in CVD Subjects

Abstract

This paper analyzes, through computational simulations, which spectral filters increase the number of discernible colors (NODC) of subjects with normal color vision, as well as red-green anomalous trichromats and dichromats. The filters are selected from a set of filters in which we have modeled spectral transmittances. With the selected filters we have carried out simulations performed using the spectral reflectances captured either by a hyperspectral camera or by a spectrometer. We have also studied the effects of these filters on color coordinates. Finally, we have simulated the results of two widely used color blindness tests: Ishihara and Farnsworth-Munsell 100 Hue (FM100). In these analyses the selected filters are compared with the commercial filters from EnChroma and VINO companies. The results show that the increase in NODC with the selected filters is not relevant. The simulation results show that none of these chosen filters help color vision deficiency (CVD) subjects to pass the set of color blindness tests studied. These results obtained using standard colorimetry support the hypothesis that the use of color filters does not cause CVDs to have a perception similar to that of a normal observer.

Is it really possible to compensate for colour blindness with a filter?

“…A thorough approach to prove this proposition could be to use simulations for a variety of individuals with CVD and all possible spectral transmittances with values of between 0 and 1. This has not been carried out yet, although Martínez‐Domingo et al have analysed a wide set of filters 21 . Another possible approach could be to use information theory 23 to prove that if the spectral responsivity of one of the three cone photoreceptors is shifted with respect to its normal peak position, this then makes it impossible to achieve an improvement in colour vision by introducing changes in the signals that the photoreceptors respond to.…”

“…The second reason is that quantifying colour vision improvements is still a challenging issue. Computations using simulated CVD vision modelling have been carried out with the following metrics: the number of discernible colours (interpreted as a measure for gamut expansion 21 ), colour differences for a given set of colour stimuli with or without the filter added 18,20 , and colour differences for subjects with CVD with or without the filter added, taking the colour perceived by a subject with normal colour vision as a reference 22 …”

“…When there are psychophysical experiments available, the possible improvement obtained with the filter has been measured by studying colour discrimination thresholds (eg, CAD test 24 ), recognition tests such as the Ishihara test, sorting tests such as Farnsworth D‐15 or FM‐100 (also used in simulations in Martínez‐Domingo et al 21 ), colour‐naming 25,26 or a subjective evaluation by the subjects 27 . Among these tests, it has been proposed that sorting tests such as the FM‐100 are more appropriate for evaluating possible improvements than recognition tests such as the Ishihara 21 . In addition, there are issues related to the way the tests are carried out; if they are displayed on a screen then the use of the filters could significantly alter the perceived colour because of the narrowness of the spectral primaries of the displays.…”

“…Computations using simulated CVD vision modelling have been carried out with the following metrics: the number of discernible colours (interpreted as a measure for gamut expansion 21 ), colour differences for a given set of colour stimuli with or without the filter added 18,20 , and colour differences for subjects with CVD with or without the filter added, taking the colour perceived by a subject with normal colour vision as a reference. 22 When there are psychophysical experiments available, the possible improvement obtained with the filter has been measured by studying colour discrimination thresholds (eg, CAD test 24 ), recognition tests such as the Ishihara test, sorting tests such as Farnsworth D-15 or FM-100 (also used in simulations in Martínez-Domingo et al 21 ), colour-naming 25,26 or a subjective evaluation by the subjects. 27 Among these tests, it has been proposed that sorting tests such as the FM-100 are more appropriate for evaluating possible improvements than recognition tests such as the Ishihara.…”

“…Many of these filters have been tested and analysed. [7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22] Considering the large variety of analyses, psychophysical experiments, simulations and number of subjects used in these papers, researchers have reached the following conclusions:…”