This paper considers a novel application of x-ray imaging of planks, for the purpose of detecting knots in high quality furniture wood. X-ray imaging allows the detection of knots invisible from the surface to conventional cameras. Our approach is based on texture analysis, or more specifically, discriminative dictionary learning. Experiments show that the knot detection and segmentation can be accurately performed by our approach. This is a promising result and can be directly applied in industrial processing of furniture wood.