This work treats cross-section (CS) segmentation in digital images of rough wood log ends. Existing CS segmentation approaches are focused on computed tomography CS images of logs and no approach and experimental evaluation for digital images has been presented so far. Segmentation of cross-sections in rough log end images is a prerequisite for the development of novel log end analysis applications (e.g. biometric log recognition or automated log grading). We propose a simple and fast computable similarity-based region growing algorithm for CS segmentation. In our experiments we evaluate different texture features (Local binary patterns & Intensity histograms) and histogram distances. Results show that the algorithm achieves the most accurate results in combination with intensity histograms and the earth movers distance. Generally, we conclude that for certain applications simple texture features and a matured distance metric can outperform higher-order texture features and basic distance metrics.