Adaptive spectral characterization model for various types of displays

In this study, an adaptive spectral characterization model is proposed to be applicable to various types of displays. The model consists of a characteristic fitting module and a crosstalk compensation module, similar to residual connections. The characteristic fitting module adaptively learns the display's tone reproduction curve during iterative training, improving the prediction accuracy when training data are limited. Combined with the innovative oblique cubic sampling method, the model requires only 128 training samples to achieve excellent performance across five types of displays, with an average spectral root mean square error of less than 10^-4 and an average CIEDE2000 color difference of less than 1.2. Additionally, with slight adjustments, the model can be applied to chromaticity characterization, effectively meeting specific application requirements.