The osteoporosis is a common public health problem that weakens bones, which increases risk of fracture, due to the porous trabecular bone microarchitecture. In a clinical routine, the basis osteoporosis screening is done by measuring the bone mineral density (BMD) determined by dual-energy X-ray absorptiometry (DXA), which is an inefficient and manually non-invasive procedure. Therefore, automatic methods that potentially characterize the trabecular bone texture using texture like analysis are deemed very relevant. Fractal is widely used for trabecular bone texture analysis, where the fractal dimension (FD) is the main component of fractal features for quantifying the complexity of the images. The aim of this article is to enhance the quality of the preprocessing prior to the application of the fractal analysis in order to yield a robust FD estimate in frequency domain. A preprocessing stage involves a combination of contrast enhancement followed by Discrete Cosine Transform (DCT), where the DCT coefficients is inputted to differential box-counting method (DBCM) to estimate the FD. This approach is shown to provide better results, in term of discriminating pathological and healthy trabecular bones, using statistical Wilcoxon rank sum test, than estimating FD without preprocessing stage. The testing and evaluation phases are carried out using the Lille INSERM U 703 database that contains a set of ROIs (Region Of Interest) CT-SCAN texture images of both healthy and pathological trabecular bones