A three-dimensional mapping method of human faces in a computer virtual world comprises the following steps: Step 1, providing a human face photo, conducting human face feature recognition by using the computer vision technology, and acquiring human face feature data; Step 2, selecting the most matching three-dimensional human face model picture of the human face feature data in a prefabricated three-dimensional human face model picture, or generating a three-dimensional human face model picture according to the human feature data; and Step 3, combining the human face photo and the thee-dimensional human face model picture. A front three-dimensional virtual human face is rapidly and accurately generated according to the human face photo, and the human face picture is perfectly matched with the three-dimensional model. Even if a user who has no three-dimensional knowledge can create a three-dimensional role of the user himself in a short time. The method can be widely applicable to personal images or virtual characters in the virtual world, such as network interaction, web2.0, a virtual platform, communities, etc.