Face detection is a topical problem in computer graphics, image processing and other areas of science. There are many face detection methods for static images but only a few methods are suitable for face detection in a dynamic environment, because they are time-consuming and their implementation often requires complicated preparation process. This work describes algorithm of face detection, which is designed to be able to perform real-time face detection in a dynamic environment. This algorithm is based on the search of face features (skin tone, contour) and is similar to face region extraction. The algorithm is based on three steps: all pixels are classified into two classes – skin and non-skin pixel classes, contour detection; merging of skin pixels in regions; resulting region analysis to determine face existence in the region. HSV colour system is used to detect skin colour because it describes skin tone much better than the RGB colour system. To determine skin pixels probability distribution of skin tone is used, which is obtained from a sample of face in dynamic environment material (example, video material). This can be done either before the beginning of the algorithm or during its operation. Such an approach of learning algorithm is simple, fast and adaptive. This algorithm is able to perform face detection in real-time with faces at different turn angles, different facial expressions, various lighting conditions, as well as with multiple faces simultaneously. Learning process of the algorithm allows adapting to environmental and lighting conditions. It means that skin tones do not need to be in certain tone values, i.e. it can vary. Developed algorithm is also able to process the image quickly; therefore, it is suitable for both static images and video, where algorithm creates small delay.