The evaluation of femur bone strength is a key component for fracture risk assessment. Recently, it has been demonstrated that bone strength depends not only on bone mass but also on factors related to bone quality, such as trabecular architecture and morphology. Current clinical methods for assessment of bone quality are largely dependent on assessing bone mass. However, these methods do not provide any information about bone structure which is considered to be an equally important factor in assessing bone quality. In this work, ridgelet transform based multi-scale geometric analysis is performed in radiographic images to characterize the trabecular structure. The trabecular regions of normal and abnormal human femur bone images (N=40) recorded under standard condition are used for the study. The regions of interest in the bone images are subjected to ridgelet transform for extracting useful features that evaluate changes taking place in the architecture of bone. The extracted features are correlated with apparent mineralization which is a key of representative architectural variation of the trabecular bone. Further to validate the results, images are also classified based on the extracted features using Support Vector Machines (SVM) with four different kernels. Results show that ridgelet transform are able to differentiate normal and abnormal femur bone images. The values of derived features such as energy and homogeneity are found to have good correlation with apparent mineralization. In abnormal images, the variations in the observed features are attributed to loss in bone mass, inhomogeneity and anisotropic nature of such images. Further, classification performed using polynomial kernel based SVM is found to be effective in terms of number of support vectors, sensitivity and specificity. Hence it appears that this method is useful for gross abnormality detection and micro-damage analysis.