Calibrating color cameras using metameric blacks

0
361

In the article titled “Calibrating color cameras using metameric blacks,” authors Alsam and Lenz present a novel methodology for calibrating color cameras through the use of “metameric blacks.” This approach is based on the premise that the difference between two metameric spectra lies in the black space of the sensor, and its primary objective is to estimate the spectral sensitivity curves of the camera sensor.

The process begins by selecting calibration color triplets and using convex linear combinations of calibration chip spectra that exhibit sensor response values both above and below. The difference between the measured reflectance spectrum and the calculated numerical metamerism resides in the sensor’s null space. For each measured spectrum, a collection of color signals within this null space is generated, and robust principal component analysis (PCA) is employed to compute projections characterizing the camera’s null space. The implementation of this methodology, tested on Nikon D70 and MegaVision cameras, proved robust against outliers and noise, outperforming traditional methods such as truncated Singular Value Decomposition (TSVD) and Tikhonov regularization (TR).

Experimental results emphasize that, using the new method, the estimation of spectral sensitivities was comparable or even superior to those obtained by conventional methods, especially in the case of data not included in the calibration. Indeed, for the calibrated cameras, the absolute difference between the measured and estimated responses was smaller using the new metamer-based method than with TSVD and TR.

The construction of “metameric blacks” is technically described, detailing how they are calculated and used to determine an orthogonal basis in the sensor space. This provides a more accurate and less noise-sensitive estimation, transferring noise variability from the response data to the spectral space. A key feature of the proposed method is its ability to assess the “goodness” of the calibration based on the dimensionality of the obtained spectral data.

Finally, the article concludes that this approach is not only efficient but also more robust compared to traditional methods, suggesting potential improvements for future research that could focus on a more detailed evaluation of noise characteristics .

Review on the publisher’s website