Contextual Multivariate Segmentation of Pork Tissue from Grating-Based Multimodal X-Ray Tomography

X-ray computed tomography is increasingly used as a nondestructive method for studying three dimensional food structures. For meat products, studies have focused mainly on fat and protein content due to limited contrast capabilities of absorption based techniques. Recent advances in X-ray imaging have made novel X-ray image modalities available, where the refraction and scattering of X-rays is obtained simultaneously with the absorption properties, providing enhanced contrast for soft biological tissues. This paper demonstrates how data obtained from grating-based imaging can be segmented by means of multivariate and contextual methods to improve the classification of soft tissues in meat products. The results show that the presented segmentation method provides improved classification over univariate segmentation.