Automatic detection of fractures from x-ray images is considered as an important process in medical image analysis by both orthopaedic and radiologic point of view. This paper proposes a fusion-classification technique for automatic fracture detection from long bones, in particular the leg bones (Tibia bones). The proposed system has four steps, namely, preprocessing, segmentation, feature extraction and bone detect ion, which uses an amalgamation of image processing techniques for successful detection of fractures. Three classifiers, Feed Forward Back Propagation Neural Networks (BPNN), Support Vector Machine Classifiers (SVM) and Naïve Bayes Classifiers (NB) are used during fusion classification. The results from various experiments prove that the proposed system is shows significant improvement in terms of detection rate and speed of classification.