Thin films loaded with the drug paracetamol were produced from polymer blends formed by hydroxypropylmethylcellulose (HPMC), polyvinylpyrrolidone (PVP) and polyethyleneglycol (PEG), at various mass ratios of polymers and drug defined by a D-optimal experimental design. NIR hyperspectral images were obtained from each thin film formulation and the pixel-to-pixel quantification of the constituents were carried out by partial least square (PLS) and multivariate curve resolution-alternating least square (MCR-ALS) with three different calibration/validation strategies. These strategies differ in the way to construct the calibration and validation matrices and they had to be carried out to suppress the bias on the quantification of the constituents in the polymer blend. The errors of prediction in the models from MCR-ALS were influenced by the calibration/validation strategy employed, but they were similar to the ones from PLS model. Concentration distribution maps were built after pixel-to-pixel predictions and their characteristics were analyzed.