imsegfmm

Binary image segmentation using fast marching method

Syntax

BW = imsegfmm(W,mask,thresh)

BW = imsegfmm(W,C,R,thresh)

BW = imsegfmm(W,C,R,P,thresh)

[BW,D] = imsegfmm(___)

Description

BW = imsegfmm(W,mask,thresh) segments a weight array W using the fast marching method. mask is a logical array that specifies seed locations. thresh specifies the threshold level that separates foreground from background pixels.

BW = imsegfmm(W,C,R,thresh) segments a 2-D image, with seed locations specified by column and row indices C and R, respectively.

BW = imsegfmm(W,C,R,P,thresh) segments a 3-D image, with seed locations specified by column, row, and plane indices C, R, and P, respectively.

[BW,D] = imsegfmm(___) also returns the normalized geodesic distance map D computed using the fast marching method. BW is a thresholded version of D, where all the pixels that have normalized geodesic distance values less than or equal to thresh are considered foreground pixels and set to true. You can obtain different segmentation results by thresholding D at different levels.

Examples

Segment Image Using Fast Marching Method Algorithm

This example shows how to segment an object in an image using Fast Marching Method based on differences in grayscale intensity as compared to the seed locations.

Read image.

I = imread('cameraman.tif');
imshow(I)
title('Original Image')

Create mask and specify seed location. You can also use roipoly to create the mask interactively.

mask = false(size(I)); 
mask(170,70) = true;

Compute the weight array based on grayscale intensity differences.

W = graydiffweight(I, mask, 'GrayDifferenceCutoff', 25);

Segment the image using the weights.

thresh = 0.01;
[BW, D] = imsegfmm(W, mask, thresh);
figure
imshow(BW)
title('Segmented Image')

You can threshold the geodesic distance matrix D using different thresholds to get different segmentation results.

figure
imshow(D)
title('Geodesic Distances')

Segment Object in Volume Based on Intensity Differences

This example segments the brain from MRI data of the human head.

Load the MRI data.

load("mri")
V = squeeze(D);

Visualize the data.

sizeO = size(V);
figure
slice(double(V),sizeO(2)/2,sizeO(1)/2,sizeO(3)/2);
shading interp
colormap("gray")
title("Original")

Set the seed locations.

seedR = 75; 
seedC = 60; 
seedP = 10;

Compute weights based on grayscale intensity differences.

W = graydiffweight(V,seedC,seedR,seedP,"GrayDifferenceCutoff",25);

Segment the image using the weights.

thresh = 0.002;
BW = imsegfmm(W,seedC,seedR,seedP,thresh);

Visualize the segmented image using an isosurface.

figure
p = patch(isosurface(double(BW)));
p.FaceColor = "red";
p.EdgeColor = "none";
daspect([1 1 27/64]);
camlight
lighting phong

Input Arguments

W — Weight array

non-negative numeric array

Weight array, specified as a non-negative numeric array. You can compute the weight array by using the graydiffweight or gradientweight functions. Large values in W identify the foreground (object) and small values identify the background.

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

mask — Seed locations mask

logical array

Seed locations mask, specified as a logical array of the same size as W. Locations where mask is true are seed locations. If you use graydiffweight to create the weight matrix W, it is recommended that you use the same value of mask with imsegfmm that you used with graydiffweight.

Data Types: logical

thresh — Threshold

number in the range [0, 1]

Threshold level used to obtain the binary image, specified as a number in the range [0, 1]. Small threshold values typically result in small foreground regions in BW, and large threshold values produce large foreground regions.

Example: 0.5

Data Types: double

C — Column index of reference pixels

numeric vector

Column index of reference pixels, specified as a numeric vector. C must contain values that are valid pixel indices in W.

Example: [50 75 93]

Data Types: double

R — Row index of reference pixels

numeric vector

Row index of reference pixels, specified as a numeric vector. R must contain values that are valid pixel indices in W.

Example: [48 71 89]

Data Types: double

P — Plane index of reference pixels

numeric vector

Plane index of reference pixels, specified as a numeric vector. P must contain values that are valid pixel indices in W.

Example: [2 4 7]

Data Types: double

Output Arguments

BW — Segmented image

logical array

Segmented image, returned as a logical array of the same size as W.

Data Types: logical

D — Normalized geodesic distance map

numeric array

Normalized geodesic distance map, returned as a numeric array of the same size as W. If W is of data type single, then D is of data type single. Otherwise, D is of data type double.

Data Types: double | single

Tips

• imsegfmm uses double-precision floating point operations for internal computations for all classes except class single. If W is of data type single, imsegfmm uses single-precision floating point operations internally.

• imsegfmm sets pixels with 0 or NaN weight values to Inf in the geodesic distance image D. These pixels are part of the background in the segmented image BW.

References

[1] Sethian, J. A. Level Set Methods and Fast Marching Methods: Evolving Interfaces in Computational Geometry, Fluid Mechanics, Computer Vision, and Materials Science, Cambridge University Press, 2nd Edition, 1999.

Extended Capabilities

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 in R2014b

R2022b: Support for thread-based environments

imsegfmm now supports thread-based environments.