Ovarian follicular monitoring is an essential diagnostic tool in obstetrics and gynecology to evaluate ovarian reserve and to estimate follicular and ovarian response to fertility treatment. Given the significant time requirement, inconvenience measuring follicles and estimating follicular development during multiple examinations, and variable results of different clinicians performing monitoring, complete automation of follicular monitoring is necessary. Computerized follicle detection is currently either semi-automated or has low performance due to limiting factors: (1) noise, (2) detecting multiple follicles very close to each other as one follicle region without finding the boundary of individual follicles, and (3) not being fast enough to be used in real-time clinical practice. To overcome these limitations, we handle noise by singular value decomposition based image compression followed by an anisotropic diffusion scheme for multiplicative speckle, and detect follicles by performing different segmentation techniques depending on features of the image (such as pixel intensity level) and features of the detected follicle areas (such as roundness). This approach allows for rapid identification and measurement of individual follicles with the ability to differentiate between the borders of adjacent follicles and the boundary between the follicle and ovarian stroma.