Segment image into foreground and background using active contours (snakes) region growing technique
The active contours technique, also called
snakes, is an iterative region-growing image segmentation
algorithm. Using the active contour algorithm, you specify initial curves on an image
and then use the
activecontour function to evolve the curves
towards object boundaries.
mask argument is a binary image that specifies the
initial state of the active contour. The boundaries of the object regions (white) in
mask define the initial contour position used for contour
evolution to segment the image. The output image
BW is a binary
image where the foreground is white (logical true) and the background is black
To obtain faster and more accurate segmentation results, specify an initial contour position that is close to the desired object boundaries.
Read and display a grayscale image.
I = imread('coins.png'); imshow(I) title('Original Image')
Specify an initial contour surrounding the objects of interest. Display the contour.
mask = zeros(size(I)); mask(25:end-25,25:end-25) = 1; imshow(mask) title('Initial Contour Location')
Segment the image by using the
activecontour function. By default, the function evolves the segmentation through 100 iterations.
bw = activecontour(I,mask);
Display the result. After 100 iterations, objects are not fully segmented from the background because the original contour is not close to the object boundaries.
imshow(bw) title('Segmented Image, 100 Iterations')
To continue evolving the segmentation, increase the number of iterations. After 300 iterations, objects are fully segmented from the background.
bw = activecontour(I,mask,300); imshow(bw) title('Segmented Image, 300 Iterations')
Read and display a grayscale image.
I = imread('toyobjects.png'); imshow(I)
r = drawrectangle;
mask = createMask(r);
Segment the image using the
'edge' method and 200 iterations.
bw = activecontour(I,mask,200,'edge');
Display the final contour over the original image in red.
hold on; visboundaries(bw,'Color','r');
Display the result of the segmentation over the original image. The object in the foreground has a blue color.
Load 3-D volumetric image data, removing the singleton dimension.
D = load('mri.mat'); A = squeeze(D.D);
Create 2-D mask for initial seed points.
seedLevel = 10; seed = A(:,:,seedLevel) > 75; figure imshow(seed)
Create an empty 3-D seed mask and put the seed points into it.
mask = zeros(size(A)); mask(:,:,seedLevel) = seed;
Perform the segmentation using active contours, specifying the seed mask.
bw = activecontour(A,mask,300);
Display the 3-D segmented image.
figure; p = patch(isosurface(double(bw))); p.FaceColor = 'red'; p.EdgeColor = 'none'; daspect([1 1 27/128]); camlight; lighting phong
A— Image to be segmented
Image to segmented, specified as a 2-D numeric matrix or 3-D numeric array.
mask— Initial contour
Initial contour at which the evolution of the segmentation begins,
specified as a binary image of the same size as
2-D and 3-D grayscale images, the size of
match the size of the image
A. For color and
mask must be a 2-D logical array
where the first two dimensions match the first two dimensions of the image
n— Maximum number of iterations
100(default) | positive integer
Maximum number of iterations to perform in evolution of the segmentation,
specified as a positive integer.
activecontour stops the
evolution of the active contour when it reaches the maximum number of
activecontour also stops the evolution if the
contour position in the current iteration is the same as the contour
position in one of the most recent five iterations.
If the initial contour position (specified by
far from the object boundaries, specify larger values of
n to achieve desired segmentation results.
method— Active contour method
comma-separated pairs of
the argument name and
Value is the corresponding value.
Name must appear inside quotes. You can specify several name and value
pair arguments in any order as
'SmoothFactor'— Degree of smoothness
Degree of smoothness or regularity of the boundaries of the segmented
regions, specified as the comma-separated pair consisting of
'SmoothFactor' and a positive number. Higher
values produce smoother region boundaries but can also smooth out finer
details. Lower values produce more irregularities (less smoothing) in
the region boundaries but allow finer details to be captured. The
default smoothness value is
0 for the
'Chan-Vese' method and
'ContractionBias'— Tendency of the contour to grow outwards or shrink inwards
Tendency of the contour to grow outwards or shrink inwards, specified as the comma-separated
pair consisting of
'ContractionBias' and a numeric
scalar. Positive values bias the contour to shrink inwards (contract).
Negative values bias the contour to grow outwards (expand). This
parameter does not guarantee that the contour contracts or expands. It
is possible that even with a positive value for this parameter, the
contour could actually expand. However, by specifying a bias, you slow
the expansion when compared to an unbiased contour. Typical values for
this parameter are between -1 and 1. The default contraction bias is
0 for the
0.3 for the
activecontour uses the boundaries of the regions in
mask as the initial state of the contour from where the
evolution starts. Holes in the mask can cause unpredictable results. Use
imfill to fill any holes in the
If a region touches the image borders, then
removes a single-pixel layer from the region, before further processing, so that
the region does not touch the image border.
To get faster and more accurate results, specify an
initial contour position that is close to the desired object boundaries,
especially for the
'edge' method, the active contour is naturally
biased towards shrinking inwards (collapsing). In the absence of any image
gradient, the active contour shrinks on its own. Conversely, with the
'Chan-Vese' method, where the contour is unbiased, the
contour is free to either shrink or expand based on the image features.
To achieve an accurate segmentation with the
specify an initial contour that lies outside the boundaries of the
object. The active contour with the
is biased to shrink, by default.
If object regions are of significantly different grayscale intensities, then
'Chan-Vese' method 
might not segment all objects in the image. For example, if the image contains
objects that are brighter than the background and some that are darker, the
'Chan-Vese' method typically segments out either the dark
or the bright objects only.
activecontour uses the Sparse-Field level-set
method, similar to the method described in ,
for implementing active contour evolution.
 T. F. Chan, L. A. Vese, Active contours without edges. IEEE Transactions on Image Processing, Volume 10, Issue 2, pp. 266-277, 2001.
 V. Caselles, R. Kimmel, G. Sapiro, Geodesic active contours. International Journal of Computer Vision, Volume 22, Issue 1, pp. 61-79, 1997.
 R. T. Whitaker, A level-set approach to 3d reconstruction from range data. International Journal of Computer Vision, Volume 29, Issue 3, pp. 203-231, 1998.