Benign Prostatic Hyperplasia (BPH) is a common old men disease for which treatment selection is very important. So it is necessary to develop tools for helping doctors to select the proper treatment. The tools based on image processing are the common methods for this purpose. The first and the most basic step of these tools is extraction the background of Prostate Biopsy (PB) images using the segmentation techniques. On the other hand there has been a big challenge to select thresholds. In literature, it was observed that the fixed threshold values or ranges, which did not always produce satisfactory results, were used for segmentation. To improve performance of segmentation of PB images with non-homogenous background, an efficient and automatic method for the extraction background of Hematoxylin and Eosin (HE) stained PB images was proposed in this study where the histogram of differences among the RGB color components was used to estimate threshold values. The developed algorithm was tested and compared with traditional methods. © 2012 Dunarea de Jos University.