Abstract
Beef is a valuable commodity in food culture. Therefore, quantitative and objective evaluation of beef has considerable attentions in agriculture and livestock industry. We are working on geometrical and topological analysis of 3D beef marbling via destructive imaging by slicing real-world beef. It is important to reduce noise, such as frozen chips generated by slicing, in the sectional images to obtain accurate analysis. In this research, we propose a novel denoising filter for 3D beef images. Our filter is based on applying a vector median filter to the color vector extrema with respect to the slicing direction. We obtain nice results compared with traditional impluse denoising filters in terms of mean squared errors.
