Search code examples
algorithmimagematlabimage-processingwavelet

What algorithm can I use to find this extracted template image in this query image?

enter image description here

I am trying to find the location of the extracted template image in the query image, I have tried using surf algorithm but it is not able to find it, please what other algorithm can i use to find this template?

Solution

  • What you are trying to do is relatively simple and straightforward, especially if you would have a clean dataset for the fingerprints.

    You can use Wavelet Transform for extracting those features.

    It would decompose your fingerprint images using a tree architecture and you can extract those desired parts of your images that you would be interested in.

    I'm pretty sure you can find many packages on GitHub.

    If you can code in Python, PyWavelets would be also a choice, otherwise you can use MATLAB:

    How to implement in MatLab:

    This code is just an example of an implemented image processing algorithm in MatLab.

    Source Code

    %WAVELET

clear all;
clc;

data=imread('sidikjari5.jpg');
r=data(:,:,1);
g=data(:,:,2);
b=data(:,:,3);

gray=double(0.333*r+0.333*g+0.333*b)/512;

[m, n] = size(gray);

%=================== WAVELET LV 1 ===================%

mm=m/2;
nn=n/2;
for i=1:mm
    for j=1:nn
        cb1=(gray(i*2-1,j*2));
        %cb1=double(cb1);
        cb2=(gray(i*2,j*2-1));
        %cb2=double(cb2);
        ca(i,j)=sqrt(cb1^2+cb2^2);
    end
end
[m1,n1]=size(ca);

for i=1:mm
    for j=1:nn
        cb1=(gray(i*2-1,j*2-1));
        cb2=(gray(i*2,j*2-1));
        cv(i,j)=(cb1+cb2)/2;
    end
end
[m2,n2]=size(cv);

for i=1:mm
    for j=1:nn
        cb1=(gray(i*2-1,j*2-1));
        cb2=(gray(i*2-1,j*2));
        ch(i,j)=(cb1+cb2)/2;
    end
end
[m3,n3]=size(ch);

for i=1:mm
    for j=1:nn
        cb1=(gray(i*2-1,j*2-1));
        cb2=(gray(i*2,j*2));
        cd(i,j)=(cb1+cb2)/2;
    end
end
[m4,n4]=size(cd);

l1=figure;
imshow(l1);
subplot(2,2,1);
imshow(ca);
title('ca');
subplot(2,2,2);
imshow(ch);
title('ch');
subplot(2,2,3);
imshow(cv);
title('cv');
subplot(2,2,4);
imshow(cd);
title('cd');

%=================== WAVELET LV 2 ===================%

mmm=m1/2;
nnn=n1/2;
for i=1:mmm
    for j=1:nnn
        cb1=(ca(i*2-1,j*2));
        cb2=(ca(i*2,j*2-1));
        ca2(i,j)=sqrt(cb1^2+cb2^2);
    end
end
[mm1,nn1]=size(ca2);

for i=1:mmm
    for j=1:nnn
        cb1=(ca(i*2-1,j*2-1));
        cb2=(ca(i*2,j*2-1));
        cv2(i,j)=(cb1+cb2)/2;
    end
end
[mm2,nn2]=size(cv2);

for i=1:mmm
    for j=1:nnn
        cb1=(ca(i*2-1,j*2-1));
        cb2=(ca(i*2-1,j*2));
        ch2(i,j)=(cb1+cb2)/2;
    end
end
[mm3,nn3]=size(ch2);

for i=1:mmm
    for j=1:nnn
        cb1=(ca(i*2-1,j*2-1));
        cb2=(ca(i*2,j*2));
        cd2(i,j)=(cb1+cb2)/2;
    end
end
[mm4,nn4]=size(cd2);

l2=figure;
imshow(l2);
subplot(2,2,1);
imshow(ca2);
title('ca2');
subplot(2,2,2);
imshow(ch2);
title('ch2');
subplot(2,2,3);
imshow(cv2);
title('cv2');
subplot(2,2,4);
imshow(cd2);
title('cd2');

%=================== WAVELET LV 3 ===================%

mmmm=mm1/2;
nnnn=nn1/2;
for i=1:mmmm
    for j=1:nnnn
        cb1=(ca2(i*2-1,j*2));
        cb2=(ca2(i*2,j*2-1));
        ca3(i,j)=sqrt(cb1^2+cb2^2);
    end
end
[mmm1,nnn1]=size(ca3);

for i=1:mmmm
    for j=1:nnnn
        cb1=(ca2(i*2-1,j*2-1));
        cb2=(ca2(i*2,j*2-1));
        cv3(i,j)=(cb1+cb2)/2;
    end
end
[mmm2,nnn2]=size(cv3);

for i=1:mmmm
    for j=1:nnnn
        cb1=(ca2(i*2-1,j*2-1));
        cb2=(ca2(i*2-1,j*2));
        ch3(i,j)=(cb1+cb2)/2;
    end
end
[mmm3,nnn3]=size(ch3);

for i=1:mmmm
    for j=1:nnnn
        cb1=(ca2(i*2-1,j*2-1));
        cb2=(ca2(i*2,j*2));
        cd3(i,j)=(cb1+cb2)/2;
    end
end
[mmm4,nnn4]=size(cd3);

level1=[ca cv;
        ch cd;];
level2ca=[ca2 cv2;
          ch2 cd2];
level3ca=[ca3 cv3;
          ch3 cd3];

l3=figure;
imshow(l3);
subplot(2,2,1);
imshow(ca3);
title('ca3');
subplot(2,2,2);
imshow(ch3);
title('ch3');
subplot(2,2,3);
imshow(cv3);
title('cv3');
subplot(2,2,4);
imshow(cd3);
title('cd3');     
 
%=================== EKSTRAKSI CIRI WAVELET LV 3 ===================%

rerataca3 = mean2(ca3); %fitur
reratach3 = mean2(ch3); %fitur
reratacv3 = mean2(cv3); %fitur
reratacd3 = mean2(cd3); %fitur

sdevca3 = std2(ca3); %fitur
sdevch3 = std2(ch3); %fitur
sdevcv3 = std2(cv3); %fitur
sdevcd3 = std2(cd3); %fitur

varca3 = sdevca3^2; %fitur
varch3 = sdevch3^2; %fitur
varcv3 = sdevcv3^2; %fitur
varcd3 = sdevcd3^2; %fitur

x=[rerataca3  reratach3 sdevca3 sdevch3  varca3  varch3]

    Update

    I would suggest to look for an already developed package on GitHub or maybe search through similar MatLab examples, if they have. Then, apply that package or example to your images to see how it would work. In this process, you may face some problems such as preprocessing that might be required to be applied on your images before you map them with any time-frequency methods.

    If not and that set a side, maybe you can start from here, if you would like to write from scratch. More specifically, Discrete Wavelet Packet Transforms might be an efficient algorithm for your images.

    Wavelet packet seems to be designed similar to trees. After you apply them to your images, you can also define and design some mathematical/statistical criteria to extract those features form your fingerprint images.