Activity 4: Rendering Color
For this activity, I used the light source emittance data and leaf reflectance data that our group obtained from the previous activities. Our reference light source is a 20W LED with a CCT of 6500K. All images were taken with a Nikon D3400 with the white balance set to daylight. Since I could not find any spectral information about my camera online, I got the spectral sensitivity data for the Nikon D70 [2] instead.
Pre-processing
Since the number of data points of the spectra were not equal, I had to interpolate some data points so that they were all equalized to the spectra with the most number of data points. Since all the intensities were relative, I then normalized all the spectra to 1.
Color rendering
The color of the leaf was rendered digitally using the light emittance spectrum
The actual and generated colors are shown in Fig. 1. We can see that the reconstructed color already captures the general color of the leaf.
Figure 1: Actual and rendered colors of a leaf.
To further verify the calculations, we attempt to reconstruct the colors from a Macbeth Color Checker [1]. The measured reflectances per patch was also provided and all computations proceed as before. However, on the first attempt, the colors looked way off. I realized that I was still using the emittance from the light source that we used in the previous activity, and that was most likely causing the error. On a hunch, I tried to search for the spectral distribution of a standard D65 illuminant. I was able to obtain spectral data from [3], and the final rendered colors are shown in Fig. 2. From these figures, we can see that the rendered colors are generally more saturated than the actual, while the gray patches are much less saturated.
Figure 1: Actual and rendered colors of a Macbeth Color Checker.
References
- M. N. Soriano, Sensors (2019).
Naemura Lab, Spectral sensitivity database (2014).
Colour and Vision Research Laboratory, Relative SPD of D65 (2007).