In this work, the application of a laser backscattering image technique as a non-destructive quality control technique for fluid food matrices was studied. The used food matrices were vegetable-based creams, which were modified according to the combination of four production factors (raw material, biopolymer type, biopolymer concentration and homogenisation system) in order to obtain a wide space of variance in terms of physicochemical properties (52 different creams). All the creams were characterised based on that imaging technique using pre-designed descriptors extracted from the captures of the generated laser patterns. The capacity to characterise creams presented by the imaging and physico-chemical data (theology and syneresis) was compared, and the effect of each production factor on their captured variance was evaluated. Both characterisations were similar. This parallelism was proved by modelling the relationship between them by carrying out regression studies. The regression coefficients were successful for most physico-chemical variables. However, the prediction of creams' properties was maximised when done over the linear combination of them all. Thus the imaging descriptors collected enough variance from the cream categories to place them according to their physicochemical properties into the generated space of physico-chemical variance. The results allowed us to conclude that this technique can be applied for the non-destructive quality control of fluid-food matrices for production processes with a wide spectrum of product categories.