New fractal features and data mining to determine food quality based on MRI

Abstract

The extraction of textural information from images to explore parameters related to food quality is very common. In this paper, the extraction of quality features from MRI is performed by a new fractal algorithm and second order statistics, as an alternative to the classical texture approaches. The proposed method needs fewer features than classical textures, computing them with a lower computational complexity. Quality characteristics from MRI of Iberian loins are extracted to validate the practical application of the proposed algorithm. The new method is compared to the standard fractal algorithm and also to the classical texture approaches. Characteristics obtained by means of the new fractal algorithm, by the standard fractal algorithm, and by the three classical texture methods are correlated to the results obtained by using physico-chemical methods. The correlations achieve coefficients higher than 0.75. Therefore, the new algorithm could be used to calculate quality parameters of meat products in a non-destructive and efficient way, being also suitable for the meat industries to characterize meat products.