Theoretical Framework to Predict Generalized Contrast-to-Noise Ratios of Photoacoustic Images With Applications to Computer Vision