uiThreshErode

Below is a demonstration of the features of the uiThreshErode function

Contents

clear; close all; clc;

Plot settings

cMap=gray(250);
faceAlpha1=1;
edgeColor1='none';

Example image of head surrounded by background

% Get a 3D image
load mri;
M=double(squeeze(D)); %example image data set
v=2./[1,1,.4]; %example voxel size

%Normalising
M=M-min(M(:));
M=M./max(M(:));

%Adding some noise
M=M+0.25.*rand(size(M));

Thresholding and erosion (and regrowing) based background removal

Start thresholding followed by dilation/erosion process

thresholdInitial=0.1; %with respect to normalised image
preBlurKernalSize=0; %with respect to normalised image
groupCropOption=0;

Uncomment to run code [L_BG]=uiThreshErode(M,thresholdInitial,preBlurKernalSize,groupCropOption);

Uncomment to run code

% % Plotting the cropped image
% logicVoxels=false(size(M));
% logicVoxels(round(size(M,1)/2),:,:)=1;
% logicVoxels(:,round(size(M,2)/2),:)=1;
% logicVoxels(:,:,round(size(M,3)/2))=1;
%
% logicVoxels1=logicVoxels;
% [F1,V1,C1]=ind2patch(logicVoxels1,M,'vb');
% [V1(:,1),V1(:,2),V1(:,3)]=im2cart(V1(:,2),V1(:,1),V1(:,3),v);
%
% logicVoxels2=logicVoxels & L_BG;
% [F2,V2,C2]=ind2patch(logicVoxels2,M,'vb');
% [V2(:,1),V2(:,2),V2(:,3)]=im2cart(V2(:,2),V2(:,1),V2(:,3),v);
%
% h1=cFigure;
%
% subplot(1,2,1);title('Original');
% xlabel('X (mm)');ylabel('Y (mm)'); zlabel('Z (mm)'); hold on;
% hp1= patch('Faces',F1,'Vertices',V1,'FaceColor','flat','CData',C1,'EdgeColor',edgeColor1,'FaceAlpha',faceAlpha1);
% axis equal; view(3); axis tight; axis vis3d; grid on;
%
% subplot(1,2,2);title('Cropped result');
% xlabel('X (mm)');ylabel('Y (mm)'); zlabel('Z (mm)'); hold on;
% hp1= patch('Faces',F2,'Vertices',V2,'FaceColor','flat','CData',C2,'EdgeColor',edgeColor1,'FaceAlpha',faceAlpha1);
% axis equal; view(3); axis tight; axis vis3d; grid on;
% colormap(cMap); colorbar;
% drawnow;

First the image data is visualized with an initial threshold applied:

Increasing the threshold (up arrow key) removes more background voxels

The threshold may also create undesired internal gaps at the level where all background is removed.

Dilations may fill in these gaps:

Comparing the final result (right) to the input (left):

GIBBON www.gibboncode.org

Kevin Mattheus Moerman, [email protected]