CODEWITHSUNDEEP
matlabimage-segmentation
user3217708
user3217708

Reputation: 45

How to find precision and recall of segmented image?

I have tried this code:

function [Fvalue,precision,recall,accuracy,JaccardIndex,TP,FP,TN,FN,FPrate,TPrate,MCC] = compareBinaryImages( reference, toTest )
%COMPAREBINARYIMAGES Compute various similarity metrics between two binary images
%   reference = grouth truth binary image
%   toTest = binary image to be compared to the reference image

if(ndims(reference)~=2 && ndims(toTest)~=2) 
    error('Inputs must be two 2-dimensional matrices'); 
end; 

%  TP = numel(find(reference==1 & toTest==1)==1); % True positive
%  FP = numel(find(reference==0 & toTest==1)==1); % False positive
%  TN = numel(find(reference==0 & toTest==0)==1); % True negative
%  FN = numel(find(reference==1 & toTest==0)==1); % False negative


 TP = nnz(reference==1 & toTest==1); % True positive
 FP = nnz(reference==0 & toTest==1); % False positive
 TN = nnz(reference==0 & toTest==0); % True negative
 FN = nnz(reference==1 & toTest==0); % False negative

P = TP + FN; % Total positive for the true class (= reference)
N = FP + TN; % TOtal negative for the true class (= reference)

FPrate = FP/N; % False positive rate
TPrate = TP/P; % True positive rate

precision = TP/(TP+FP);
recall = TP/P;
accuracy = (TP+TN)/(P+N);

MCC = (TP*TN-FP*FN)/sqrt((TP+FP)*(TP+FN)*(TN+FP)*(TN+FN));
2/((1/precision)+(1/recall));

% Alternative form for Fvalue
2*(precision*recall/(precision+recall));

% Avoid getting a division by 0 if only negatives and perfect detection
if(TN==numel(reference))
    'gaga'
    Fvalue = 1;
    warning('FValue was set to 1 as all pixels were true negatives');
else
    Fvalue=2*TP/(FP+TP+P);
end

2*TP/((FP+TP)+(TP+FN));

JaccardIndex = TP / (FP+TP+FN);

end

My testing image is enter image description here

My reference image is enter image description here

But I got the result as NaN. What's wrong with this?

Upvotes: 0

Views: 1882

Answers (1)

user2999345
user2999345

Reputation: 4195

the attached function, compareBinaryImages, expects binary inputs, i.e., input images which contain only 0's and 1's. binarize your images and apply this function to the binary images results in several numeric outputs, non of which is Nan:

testingImg = imread('testing.png');
referenceImg = imread('reference.png');
% grayscale
testingGray = rgb2gray(testingImg);
referenceGray = rgb2gray(referenceImg);
% binarize
testingBw = testingGray == 255;
referenceBw = referenceGray == 255;
% compare *binary* images
[Fvalue,precision,recall,accuracy,JaccardIndex,TP,FP,TN,FN,FPrate,TPrate,MCC] = ...
    compareBinaryImages(referenceBw, testingBw);

Upvotes: 1

Related Questions