Lower vertebrates, including fish, can rapidly alter skin lightness through changes in melanin concentration and melanosomes’ mobility according to various factors, which include background color, light intensity, ambient temperature, social context, husbandry practices and acute or chronic stressful stimuli. Within this framework, the determination of skin chromaticity parameters in fish species is estimated either in specific areas using colorimeters or at the whole animal level using image processing and analysis software. Nevertheless, the accurate quantification of melanin content or melanophore coverage in fish skin is quite challenging as a result of the laborious chemical analysis and the typical application of simple optical imaging methods, requiring also to euthanize the fish in order to obtain large skin samples for relevant investigations. Here we present the application of a novel hybrid confocal fluorescence and photoacoustic microscopy prototype for the label-free imaging and quantification of melanin in fish scales samples with high spatial resolution, sensitivity and detection specificity. The hybrid images are automatically processed through optimized algorithms, aiming at the accurate and rapid extraction of various melanin accumulation indices in large datasets (i.e., total melanin content, melanophores’ area, density and coverage) corresponding to different fish species and groups. Furthermore, convolutional neural network-based algorithms have been trained using the recorded data towards the classification of different scales’ samples with high accuracy. In this context, we demonstrate that the proposed methodology may increase substantially the precision, as well as, simplify and expedite the relevant procedures for the quantification of melanin content in marine organisms.