Breast cancer is one of the most dangerous diseases among women, and is the second lethal one among cancers. Mammography, as the oldest kind of imaging of the breast tissue, is one of the most prevalent methods used to detect breast cancer. One of the most regular symptoms of breast cancer is the exixtence of microcalcifictions in this tissue. Hence, detection of this symptom in mammograms, compared to other symptoms is more valuable. If microcalcifications con not be traced as elusters, because of being very smal or having different patterns, detecting them will be very diffivult.Due to the large number of patients,eror and fatigue of radiologist's eyes or lack of their experience, these symptoms are missed in 20 to 30 perrent of cases. The aim of thisproject is to present a method, based on image analysis, to detect microcalcifications in mammograms with reasonable speed and accuracy. Our method consists of tee steps: pre-processing and selecting some pixels as candidates for microcalcification, extracting features from these pixels and classifyinf these extracted features. We used the Contourlet tansform in the pre-processing part. Applying this pre-rocessing on the images results to a better contrast compared to the main image. In the next step, to quantify the extracted features, we tried to use some features that radiologist use when they want to analyse a mammogram. After the pre-processing step morpohological features are used to extract features, which are based on grey scale and tezture features that consist of local binary patterns. In the last step , neural network and support vector machine are used as classifers, and the images were reconstructed after applying them. To evalute the proposed method, the results of the proposed algoritms are compared with the diagnosis of the radiologist using two criteria. The final result of detecting microcalcifications was 82/84% in the accuracy circles and 80/17% in the whole images. The best result was derived when SVM was used as a classifier.