watershed

Watershed transform

Syntax

L = watershed(A)

L = watershed(A,conn)

Description

The watershed transform finds "catchment basins" or "watershed ridge lines" in an image by treating it as a surface where light pixels represent high elevations and dark pixels represent low elevations. The watershed transform can be used to segment contiguous regions of interest into distinct objects.

L = watershed(A) returns a label matrix L that identifies the watershed regions of the input matrix A.

L = watershed(A,conn) specifies the connectivity to be used in the watershed computation.

Examples

Compute Watershed Transform and Display Resulting Label Matrix

Create a binary image containing two overlapping circular objects. Display the image.

center1 = -40;
center2 = -center1;
dist = sqrt(2*(2*center1)^2);
radius = dist/2 * 1.4;
lims = [floor(center1-1.2*radius) ceil(center2+1.2*radius)];
[x,y] = meshgrid(lims(1):lims(2));
bw1 = sqrt((x-center1).^2 + (y-center1).^2) <= radius;
bw2 = sqrt((x-center2).^2 + (y-center2).^2) <= radius;
bw = bw1 | bw2;
imshow(bw)
title('Binary Image with Overlapping Objects')

Calculate the distance transform of the complement of the binary image. The value of each pixel in the output image is the distance between that pixel and the nearest nonzero pixel of bw.

D = bwdist(~bw);
imshow(D,[])
title('Distance Transform of Binary Image')

Take the complement of the distance transformed image so that light pixels represent high elevations and dark pixels represent low elevations for the watershed transform.

D = -D;
imshow(D,[])
title('Complement of Distance Transform')

Calculate the watershed transform. Set pixels that are outside the ROI to 0.

L = watershed(D);
L(~bw) = 0;

Display the resulting label matrix as an RGB image.

rgb = label2rgb(L,'jet',[.5 .5 .5]);
imshow(rgb)
title('Watershed Transform')

Compute Watershed Transform of 3-D Binary Image

Make a 3-D binary image containing two overlapping spheres.

center1 = -10;
center2 = -center1;
dist = sqrt(3*(2*center1)^2);
radius = dist/2 * 1.4;
lims = [floor(center1-1.2*radius) ceil(center2+1.2*radius)];
[x,y,z] = meshgrid(lims(1):lims(2));
bw1 = sqrt((x-center1).^2 + (y-center1).^2 + ...
           (z-center1).^2) <= radius;
bw2 = sqrt((x-center2).^2 + (y-center2).^2 + ...
           (z-center2).^2) <= radius;
bw = bw1 | bw2;
figure, isosurface(x,y,z,bw,0.5), axis equal, title('BW')
xlabel x, ylabel y, zlabel z
xlim(lims), ylim(lims), zlim(lims)
view(3), camlight, lighting gouraud

Compute the distance transform.

D = bwdist(~bw);
figure, isosurface(x,y,z,D,radius/2), axis equal
title('Isosurface of distance transform')
xlabel x, ylabel y, zlabel z
xlim(lims), ylim(lims), zlim(lims)
view(3), camlight, lighting gouraud

Complement the distance transform, force nonobject pixels to be Inf, and then compute the watershed transform.

D = -D;
D(~bw) = Inf;
L = watershed(D);
L(~bw) = 0;
figure
isosurface(x,y,z,L==1,0.5)
isosurface(x,y,z,L==2,0.5)
axis equal
title('Segmented objects')
xlabel x, ylabel y, zlabel z
xlim(lims), ylim(lims), zlim(lims)
view(3), camlight, lighting gouraud

Improve Watershed Segmentation Using H-Minima Transform

You can suppress shallow regional minima to avoid oversegmentation during watershed segmentation.

Load an RGB image of pears to segment. Convert the image to grayscale and display it. The center of each pear is bright, corresponding to a regional maximum.

RGB = imread("pears.png");
I = im2gray(RGB);
imshow(I)

In watershed segmentation, the image is analogous to a surface comprised of watershed lines and catchment basins. When water flows into the surface, it pools in the catchment basins. In a grayscale image, local minima are the catchment basins. To segment the pears, invert the image so the centers of the pears become regional minima.

Icomp = imcomplement(I);
imshow(Icomp)

Display the inverted image as a 3-D surface, in which the third dimension for each pixel is its intensity value. The deeper regions for each pear have spiky bottoms, indicating many shallow regional minima, like catchment basins into which water can pool.

surf(Icomp,EdgeColor="none")
colormap(gray)

Segment the unfiltered image and display the result as a label overlay. The image is oversegmented, meaning there are many small masks rather than one mask for each pear.

L = watershed(Icomp);
overlay = labeloverlay(I,L);
imshow(overlay)

Suppress the shallow minima by applying the H-minima transform. The value for h has been determined by using trial and error. Change the value to see how the h value affects the segmentation result.

h = 30;
Ifilt = imhmin(Icomp,h);

Display the filtered image as a 3-D surface.

surf(Ifilt,EdgeColor="none")
colormap(gray)

Segment the filtered image and display the result. The image contains approximately one mask for each pear in the foreground.

Lfilt = watershed(Ifilt);
overlayfilt = labeloverlay(I,Lfilt);
imshow(overlayfilt)

Input Arguments

A — Input image

numeric array | logical array

Input image, specified as a numeric or logical array of any dimension.

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64 | logical

conn — Pixel connectivity

4 | 8 | 6 | 18 | 26 | 3-by-3-by- ... -by-3 matrix of 0s and 1s

Pixel connectivity, specified as one of the values in this table. The default connectivity is 8 for 2-D images, and 26 for 3-D images.

Value	Meaning
Two-Dimensional Connectivities
4	Pixels are connected if their edges touch. The neighborhood of a pixel are the adjacent pixels in the horizontal or vertical direction.	Current pixel is shown in gray.
8	Pixels are connected if their edges or corners touch. The neighborhood of a pixel are the adjacent pixels in the horizontal, vertical, or diagonal direction.	Current pixel is shown in gray.
Three-Dimensional Connectivities
6	Pixels are connected if their faces touch. The neighborhood of a pixel are the adjacent pixels in: One of these directions: in, out, left, right, up, and down	Current pixel is shown in gray.
18	Pixels are connected if their faces or edges touch. The neighborhood of a pixel are the adjacent pixels in: One of these directions: in, out, left, right, up, and down A combination of two directions, such as right-down or in-up	Current pixel is center of cube.
26	Pixels are connected if their faces, edges, or corners touch. The neighborhood of a pixel are the adjacent pixels in: One of these directions: in, out, left, right, up, and down A combination of two directions, such as right-down or in-up A combination of three directions, such as in-right-up or in-left-down	Current pixel is center of cube.

For higher dimensions, watershed uses the default value conndef(ndims(A),"maximal").

Connectivity can also be defined in a more general way for any dimension by specifying a 3-by-3-by- ... -by-3 matrix of 0s and 1s. The 1-valued elements define neighborhood locations relative to the center element of conn. Note that conn must be symmetric about its center element. See Specifying Custom Connectivities for more information.

Note

If you specify a nondefault connectivity, then pixels on the edge of the image might not be considered to be border pixels. For example, if conn = [0 0 0; 1 1 1; 0 0 0], elements on the first and last row are not considered to be border pixels because, according to that connectivity definition, they are not connected to the region outside the image.

Data Types: double | logical

Output Arguments

L — Label matrix

numeric array of nonnegative integers

Label matrix, specified as a numeric array of nonnegative integers. The elements labeled 0 do not belong to a unique watershed region. The elements labeled 1 belong to the first watershed region, the elements labeled 2 belong to the second watershed region, and so on.

Tips

• To prevent oversegmentation, remove shallow minima from the image by using the imhmin function before you use the watershed function.

Algorithms

watershed uses the Fernand Meyer algorithm [1].

References

[1] Meyer, Fernand, "Topographic distance and watershed lines,” Signal Processing , Vol. 38, July 1994, pp. 113-125.

Extended Capabilities

C/C++ Code Generation

Generate C and C++ code using MATLAB® Coder™.

Usage notes and limitations:

• watershed supports the generation of C code (requires MATLAB^® Coder™). Note that if you choose the generic MATLAB Host Computer target platform, watershed generates code that uses a precompiled, platform-specific shared library. Use of a shared library preserves performance optimizations but limits the target platforms for which code can be generated. For more information, see Types of Code Generation Support in Image Processing Toolbox.

• Supports only 2-D images

• Supports only 4 or 8 connectivity

• Supports images containing up to 65,535 regions

• Output type is always uint16

GPU Code Generation

Generate CUDA® code for NVIDIA® GPUs using GPU Coder™.

Usage notes and limitations:

• Supports only 2-D images

• Supports only 4 or 8 connectivity

• Supports images containing up to 65,535 regions

• Output type is always uint16

Thread-Based Environment

Run code in the background using MATLAB® backgroundPool or accelerate code with Parallel Computing Toolbox™ ThreadPool.

This function fully supports thread-based environments. For more information, see Run MATLAB Functions in Thread-Based Environment.

Version History

Introduced before R2006a

R2022b: Support for thread-based environments

watershed now supports thread-based environments.