The plant carotenoid (Car) content plays a crucial role in the xanthophyll cycle and provides essential information on the physiological adaptations of plants to environmental stress. Spectroscopy data are essential for the nondestructive prediction of Car and other traits. However, Car content estimation is still behind in terms of accuracy compared to other pigments, such as chlorophyll (Chl). Here, I examined the potential of using the continuous wavelet transform (CWT) on leaf reflectance data to create vegetation indices (VIs). I compared six CWT mother families and six scales and selected the best overall dataset using random forest (RF) regressions. Using a brute-force approach, I created wavelet-based VIs on the best mother family and compared them against established Car reflectance-based VIs. I found that wavelet-based indices have high linear sensitivity to the Car content, contrary to typical nonlinear relationships depicted by the reflectance-based VIs. These relations were theoretically contrasted with the synthetic data created using the PROSPECT-D radiative transfer model. However, the best selection of wavelength bands in wavelet-based VIs varies greatly depending on the spectral characteristics of the input data before the transformation.