Report of project1 Introduction to the computer vision
Xianglong Feng
Active Contour snakes
In computer vision, it is a very important part to recognizing objects. This approach is usually depends
on identifying particular shapes in an image. The area of shape representation is concerned with finding
ways of describing shape that are sufficiently general to be useful for a range of objects, whilst at the
same time allowing their computation from image data, and facilitating comparisons of similar shapes.
There have been many algorithms to find the edge of the shape while the results are usually not
showing an entire block for the edge is usually broken up.
The contour that we are looking for should be smooth and closed, and what is the most important thing
is that the outline should stay close to the edge of the shape. The main idea of the algorithm is that we
draw a large closed area with many points. The big area is supposed to include the shape that we want
to recognize. Then the area will shrink with help of the algorithm and top at the edge of the shape.
When the area stops shrink we will get the target area that covers the shape we want.
There are three forces that will change the size of area which individually make the area shrinks, have a
smooth boundary and stop at the edge of the target shape. The first force is to make the distance
between the points shorter so that the area will shrink. The second force is to help the keep the
curvature of each point to the minimum so that the edge of the area is smooth. The last force finds the
edge of the shape and so decides when the shrinking progress should end. With the three forces, the
area shrinks and stops at the edge of the target edge.
Outline of the algorithmic approach
This algorithm circles the area by moving the selected points to the edge of the shape and links all the
points to an outline. The points are moved with three forces. The forces are calculated as Eg, Ec, and Es.
Eg (usually defined as -Eg)is related to the edge of the shape and can be got from the gradient of the
image. The gradient magnitude of the edge is much greater. The Ec is related to the distance of each
point. Try to reduce the Ec will make the circle shrink to the shape. Es is related to the curvature of each
point. The smaller of the curvature the smooth of the circle since most of the edge is smooth. Given the
corner of the shape, the curvature of that point could be ignored.
With the three forces the energy function could be developed with E=aEc+bEs+cEg. We could calculate
the energy function in the local area which center is the current point i (Pi), such that we can find a
minimum E in it. And the position of the minimum E in the local area becomes the position that Pi need
moved to.
This progress will be iterated until most of the points stop moving.
Just as the algorithm descript, I developed the software as the flow chart shows.
