Color Enhancement by Optimizing the Illumination Spectrum

2.1 Synthesis and control of illumination using LEDs

The SPD of illumination can be represented as a linear combination of the SPDs of a monochrome LED. Let us consider a lighting system consisting of N-colored LEDs. The SPD of objective illuminant I_obj(λ) is described as:

where λ is wavelength and w_i and e_i(λ) are the weight and SPD of the i th color LED of the lighting system. An object’s color is enhanced by changing the values of w = [w₁, ..., w_N]. To maintain the color balance of the object under illumination before the SPD of the illumination is changed, let us consider the constraint conditions where the targeted illumination satisfies metameric^*1 white. This means that chromaticity^*2 values of standard white and its luminance are maintained after the SPD of illumination is changed. These conditions are described as:

where I_org(λ) is the SPD of the original illumination. Here, f_L(I(λ), r(λ)), f_a(I(λ), r(λ)), and f_b(I(λ), r(λ)) are functions for calculating CIE-Lab^*3 values, where I(λ) is the SPD of illumination and r(λ) is the spectral reflectance of an object’s surface.

2.2 Design of SPD of illumination using color chart

Let us consider the case in which a target color for enhancement is a color patch whose spectral reflectance is r_obj(λ). The SPD of the objective illumination I_obj(λ) is obtained by determining weight w = [w₁, ..., w_N], which fulfills Eqs. (2) and (3) and maximizes ε.

Let the number of target colors be represented as C. Then ε is rewritten as:

When the target colors are blue, green, and red, ε is represented as:

3.1 Experimental setup

We experimented with a 16-color LED lighting system (Telelumen Light Replicator, TeleLumen LCC) as the light source. The SPDs of each LED are shown in Fig. 1; the intensity of light for each color can be modulated. We designed the SPD of illumination using three color patches of the X-Rite ColorChecker^TM (no. 13 for blue, no. 14 for green, and no. 15 for red); the spectral reflectance of these patches is shown in Fig. 2.

Fig. 1. Spectral power distributions of each LED in the 16-color LED lighting system.

Fig. 2. Spectral reflectance of blue, green, and red patches.

We experimented with synthesizing daylight using the lighting system as the reference illumination. First, we used the color chart to design the SPD of the illumination and focused on enhancing the blue, green, and red colors. The weights of each LED were determined using the generalized reduced gradient method. Next, we evaluated the color enhancement results on a chromaticity diagram using the color chart and several old woodblock prints (ukiyo-e), which were discolored with degraded color saturation.

3.2 Results of designing SPD of illumination and color enhancement

The SPDs of daylight (blue dashed line) and the designed illumination (red solid line) are shown in Fig. 3. The SPD of the designed illumination has three peaks whose center wavelengths are 435, 530, and 634 nm. These center wavelengths correspond to the peak wavelengths of the spectral reflectance of the blue and green patches and to the rising wavelength of the spectral reflectance of the red patch.

Fig. 3. Spectral power distribution of illumination.

Images of the color chart captured under synthesized daylight and under the designed illumination are shown in Fig. 4. A comparison of color patches along the bottom row indicates that the gradation of white to gray is not different after changing the illumination. However, the saturations of other color patches were enhanced after the illumination changes, especially for the blue, green, pink, orange, and red patches. The color balance of the entire image was also maintained after the illumination change.

Fig. 4. Captured images of color chart under synthesized daylight (left) and under the designed illumination (right).

Chromaticity values of white, blue, green, and red patches plotted on the CIE-u’v’ chromaticity diagram^*4 are shown in Fig. 5. The blue diamonds and the dashed line represent colors under the illumination before the SPD was changed, and the red circles and solid line represent colors after the illumination was changed. This diagram shows that u’v’ values of white under each kind of illumination are plotted in the same spot; however, the color saturation of the other three color patches is enhanced.

Fig. 5. Color distribution of white, blue, green, and red patches plotted on CIE-u’v’ chromaticity diagram.

Images of the woodblock print (ukiyo-e) captured under daylight (left) and under the designed illumination (right) are shown in Fig. 6. In the upper images, the reddish and green parts on the clothes under the designed illumination look more vivid than those under daylight. The pink blossoms also look more reddish, and it is easier to distinguish their shapes. Likewise, the reddish and blue-green parts on the bottom prints are more vivid under the designed illumination. Note that the total color balance of these prints is properly maintained through the SPD illumination change, and the color saturation of some areas on these prints was enhanced.

Fig. 6. Observed images under daylight (left) and under the designed illumination (right).

The colors of the blue, green, and reddish parts on these prints plotted on the CIE-u’v’ chromaticity diagram are shown in Fig. 7. The circle represents a white point, and the red and blue arrows represent directions of color shifts that were enhanced by changing the SPD of illumination. The color shifts in the diagram also indicate how much the colors represented in the woodblock prints were enhanced.

Fig. 7. Color shift on CIE-u’v’ chromaticity diagram.