The present thesis is on segmentation and classification of biological objects using statistical methods. It is based on case studies dealing with different kinds of pork meat images, and we introduce appropriate statistical methods to solve the tasks in the case studies. The case studies concern classification of back bacon slices from images of back bacon, prediction of ham weight from images of the carcass, and estimation of meat percent from cross-sectional images of the carcass. The first case study investigates three different classifiers ability to classify the quality of back bacon slices. The back bacon slices are classified into four ordered classes representing the quality, and we use the sizes of different meat and fat areas of the slices as variables. The classifiers are Bayesian discriminant functions, Classification and Regression Trees, and feed-forward neural networks with back-propagation. We compare the classifiers in respect to the way they classify the back bacon slices. We predict the ham weight on pork carcasses, before the carcasses are divided into front, middle, and ham part, in the second case study. Principal component analysis applied to the shapes of the carcass is used to predict the ham weight. Given two images of a pork carcass the shapes are defined by landmarks based on anatomical knowledge. We explore how the principal components describe the variation on the carcass shapes. In the third case study we measure the sizes of meat and fat areas on cross-sectional cuts from pork carcasses and use them to estimate the meat percent. We describe a bimodal histogram transformation which is used to equalize the cross-sectional images. The meat and fat areas are segmented using deformable templates. The deformable templates are studied thoroughly, and the segmentation procedures are designed to the specific task.