Precise characterisation of photocathode mean transverse energy is critical for optimising electron beam quality. This paper presents a physics-informed image processing pipeline using Transverse Energy Spread Spectrometer data (231–291 nm), incorporating Gaussian PSF fitting, Wiener deconvolution, resolution equalisation, and noise-aware augmentation. A high-fidelity dataset of 6500 synthetic images was generated, achieving average SSIM = 0. 997 and R² 0. 98, enabling robust MTE prediction and supporting future ML-based diagnostics for next-generation photoinjectors.
Malhotra et al. (Tue,) studied this question.