End-to-end color printer calibration by total least squares regression

Area of Science:

• Color Science
• Image Processing
• Computational Imaging

Background:

• Neugebauer modeling is crucial for accurate color printer characterization.
• Traditional Least Squares (LS) methods have limitations in printer modeling.
• Accurate modeling is essential for effective halftone color printer calibration.

Purpose of the Study:

• To propose and evaluate Total Least Squares (TLS) regression for Neugebauer model parameter estimation.
• To compare the performance of TLS against traditional LS methods in printer modeling.
• To demonstrate the physical appropriateness of TLS for accounting for measurement errors.

Main Methods:

• Development of a TLS method using single-colorant step-wedge print measurements.
• Extension of the TLS method to multicolorant prints via an iterative algorithm.
• Comparison of LS and TLS techniques on two different color printers with varying halftone screen configurations.

Main Results:

• TLS methods consistently and significantly outperform LS-based techniques for parameter estimation.
• The improvement from TLS is particularly notable when the available measurement data is limited.
• TLS provides a more physically appropriate model by accounting for errors in reflectance measurements.

Conclusions:

• Total Least Squares regression offers a superior approach to Neugebauer model parameter estimation.
• The proposed TLS methods enhance the accuracy of end-to-end device characterization profiles for color printers.
• This work provides a more robust and accurate method for color printer calibration, especially in data-scarce scenarios.