The paper presents an approach for rapid and non-destructive estimation of some parameters, related to honey quality, using information, obtained from hyperspectral images. A total of 29 honey samples are investigated, representing honey from different regions in Bulgaria and from different plants. They are preliminary tested in a certified laboratory for presence of heavy metals—arsenic (As), cadmium (Cd), lead (Pb) and iron (Fe). In addition, honey samples are also tested for amount of pH, reducing sugars, sweet di-saccharide and water content. After that the hyperspectral images of the samples are obtained in the NIR region (900÷1700 nm). The data from hyperspectral image analysis is then used for creation of regression models to predict the above-mentioned honey quality parameters. For that purpose, several methods: Stochastic Gradient Descent, multi-layer perceptron, PLS, AdaBoost, Linear Regression, Gradient Boosting, Random Forest, Decision Tree and kNN are applied. The preliminary investigations show that the approach for honey quality assessment based on regression models and data, derived from hyperspectral image analysis, has promising potential for accessing some of the quality parameters.