Breast cancer has been a growing problem for women around the world. The correct interpretation of mammographic images is important for the diagnosis of breast cancer. However, this is a difficult task even for a specialist. Image processing is used to make the diagnosis less susceptible to errors. Thus, the present work proposes a new method for the search of lesion candidates in mammographic images. To verify the efficiency of the approach, the behavior of the SURF, SIFT, BRISK and ORB algorithms were analyzed, as well as the Selective Search algorithm for candidate selection. A total of 1210 mammography images were used, from the CBIS-DDSM database. Results show that the SURF algorithm presented better performance, generating on average, for each image, 4.11 candidates considered in the internal area of the lesion, reducing exploratory space by 72%, whereas the ORB generated on average 1.6 candidates with a reduction rate of 96.30%.