Active Contour Model (Snake) For Face Contour Extraction

Face contour extraction is a significant precondition of facial image analysis such as facial features detection and face recognition. Because of the variety and complexity of facial shape, extracting contour with a rigidity model is difficult and a boundary resulted from classical edge detection operators is discrete. Since snake model may be used to extract a variety of contours, it was used to overcome the above shortcomings by us.The development of active contour models results from the work of Kass, Witkin and Terzopoulos, and they offer a solution to a variety of tasks in image analysis. Active contour models may be used in image segmentation and understanding, and are also suitable for analysis of dynamic image data. The active contour model, or snake, is defined as an energy-minimizingspline-the snake's energy depends on its shape and location within theimage. Local minima of this energy then correspond to desired image properties. Snakes may be understood as a special case of a more general technique of matching a deformable model to an image by means of energy minimization. Snakes do not solve the entire problem of finding contours in image; rather, they depend on other mechanisms such as interaction with a user, interaction with some higher-level image understanding process, or information from image data adjacent in time or space. This interaction must specify an approximate shape and string position for the snake somewhere near the desired contour. A prior information is then used to push the snake toward anappropriate solution. Unlike most other image models, the snake is active, always minimizing its energy functional, therefore exhibiting dynamic behavior.Based on known research results, snake algorithm was designed and an application was developed via active contour model. Finally initial experimental results were made. In this paper, a method using a modified snake model for face contour extraction was proposed. Firstly the possible face region was found out according to heuristic rules in face profile projections. According as the prior information of face shape, snake was placed in an interesting image region and was regarded as an initial evolving curve. Then the accurate face contour was extracted via greedy snake model. Our experiment results showed that accurate and rapid extraction of face contour may be achieved via the approach prior to classical edge detection techniques. But now shortcomings are: (1)snake is sensitive to an initial position and must be placed in an interesting region someway; it may converge at a local extremum point or even diverge; (3)it is sensitive to light too.