Cover Letter 1. Examples of the Opponent Color Spectral Sensitivities Obtained by the Proposed Method

This is the cover letter for our paper entitled “A Tone Reproduction Operator for All Luminance Ranges Considering Human Color Perception”. This document explains the details of the proposed method, especially, the opponent color spectral sensitivities computed during the operation of the proposed method and the wavelength discrimination curves that correspond to the adapted retinal illuminance levels. In addition, we have made a movie that shows the difference between the proposed method and other existing algorithms, and we also show some additional results. 1. Examples of The Opponent Color Spectral Sensitivities Obtained by The Proposed Method Our method computes the opponent color spectral sensitivities that satisfy the wavelength discrimination in the adapted retinal illuminance levels using the parameter functions (Paragraph 3.1 in the paper). Figure 1(a)-(c), are the opponent color sensitivities when the adapted retinal illuminance revels are 150 [Td], 8.5 [Td] and 0.85 [Td], respectively. v(λ, I), org(λ), oyg(λ) are the achromatic, the red-green opponent color and yellow-blue opponent color spectral sensitivities, respectively. The negative sensitivity of the red-green opponent color sensitivity is green sensitivity, and that of the yellow-blue sensitivity is blue sensitivity. Through Figure 1(a) to (c), the red-green opponent color spectral sensitivity decreases. In the case when the adapted retinal illuminance is 0.85 [Td] (Figure 1(c)), the yellow-blue opponent color sensitivity is higher than the redgreen opponent color spectral sensitivity. These means that the spectral sensitivity shifts toward the short wavelengths and the perception of blue increases according to the adapted light levels. These features also appear in the wavelength discrimination curves for three different retinal illuminance levels (150, 8.5, 0.85 [Td]) in Figure 2. 2. A Movie Comparing Different Methods We applied the proposed method to a sequence of photos of a scene taken at different time periods from night to morning. The photos include scotopic, mesopic, and photopic vision conditions. For comparison, we used three alternative methods; those Figure 1: Examples of opponent color spectral sensitivities obtained by applying parameter functions for three different retinal illuminance levels. of Kirk and O’Brien [KO11], Mikamo et al. [MSTK09], and Reinhard et al. [RSSF02] for scotopic, mesopic, and photopic vision, respectively. We applied those operators to the corresponding vision conditions. The range within which each of the the previously proposed methods is applied was set by the authors subjectively. The movie consists of two parts. In the first part, we show results obtained using our method and a combination of the other three previously proposed method. In the second part, c ⃝ The Eurographics Association 2014. M. Mikamo, B. Raytchev, T. Tamaki & K. Kaneda / Cover Letter Figure 2: Wavelength discrimination curves obtained by the proposed method. we show results obtained by all four methods when operating in all luminance ranges. The movie shows that the proposed method can switch smoothly between the three visual conditions with a continuous color shift. The inputs were spectral images obtained fromHDRRGB images using a method based on [Smi99]. The method in [KO11] also uses a spectral image as input, while the other two methods [MSTK09] and [RSSF02] do not. Therefore we modified them to be able to use spectral images. That is, we use CIE XYZ color matching functions to get XYZ tristimulus. After the conversion, we apply [MSTK09] and [RSSF02] to obtain the corresponding results. First, we show the results obtained by the proposed method vs. a combination of the three previously proposed method. The proposed method is able to switch smoothly between the three visual conditions, while the combination of alternative methods result in an obvious discontinuity at the points where the methods are being switched. In the second part of the movie, we applied all four method to all luminance ranges. Figure 3 shows several frames from the movie. From top to bottom, the frames show results corresponding to (a) scotopic, (b) mesopic, and (c) photopic vision conditions, respectively. Each frame contains the results obtained by the proposed method and the method from [KO11] (top row) and the methods from [MSTK09] and [RSSF02] (bottom row). The parameters for these methods are kept fixed. It can be seen that our method is able to obtain bluish image for the scotopic vision conditions, similar to the method in [KO11]. On the other hand, this color shift hardly happens for the images when the method in [MSTK09] has been used. The high “key” setting of the method from [RSSF02] gives an impression of brightness, even though this is a night scene.(Figure 3(a)) For mesopic vision, the images obtained by the proposed method have still bluish result, but not so deep as for scotopic vision (Figure 3(b)). For photopic vision, our results are almost indistinguishable from those in [RSSF02] (Figure 3(c)). Also, we applied our method to a dark scene (Figure 4, bottom). The original image was taken in the dim-light condition [Fai08]. Our method brings a change of colors on the grass and road, and the orange light near the entrance wall becomes yellowish. Our operator is able to shift the colors automatically, which is not true for the operator from [RSSF02].