How can I design a correlator to detect the character ‘G’ from the google image?

I want to design a correlator to detect the character 'G' from the given image. So can you guys please help me?

Answer 1

[Image Analyst]

If you know the G is there, you can get a mask for the G letter with the following code, which assumes the G will be the leftmost letter in the image:

rgbImage = imread('google.jpeg');
mask = ~imbinarize(rgb2gray(rgbImage)); % Get letters as foreground.
imshow(mask);
labeledImage = bwlabel(mask);  % Give an ID# to each blob.
g = ismember(labeledImage, 1); % Extract blob #1, which will be the G
imshow(g);

If you're not sure if the G is there, then you can use the normalized cross correlation (as in my other answer) to see if it's there or not.

Answer 2

[Mahesh Taparia]

Hi

The letter 'G' in Google appear to be bigger than the rest of the letter. So 'G' can be selected based on its size. You can use regionprops function to find the bounding box coordinate with maximum area and then draw rectangle on it. For example consider the code below:

a=imread('google.jpeg');

b=rgb2gray(a)<200;
c=regionprops(b,'Area','BoundingBox');
area=zeros(length(c),1);

for i=1:length(c)
    area(i)=c(i).Area;
end

[val idx]= max(area);

bbox = c(idx).BoundingBox;

imshow(a);
hold on;
rectangle('Position',bbox)

Hope it helps!

Answer 3

[Image Analyst]

If you have the G as a template, you can use normalized cross correlation with the function normxcorr2(). See attached example.

Answer 4

[yanqi liu]

sir，please check the follow code to get some information

for more information,you can read the book 《计算机视觉与深度学习实战》，Thank you！

clc;
close all;
clear all;
img = imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/757079/image.jpeg');
% get chars
hsv = rgb2hsv(img);
s = mat2gray(hsv(:,:,2));
s = imcomplement(s);
bw = im2bw(s, graythresh(s));
bw = ~bw;
figure; imshow(bw);

% gen G block
figure('Color', 'w'); hold on; axis off;
text(0.5,0.5,'G','FontSize',64);
f = getframe(gcf);

f = frame2im(f);
bw2 = ~im2bw(f);
[r,c] = find(bw2);
bw2 = bw2(min(r):max(r),min(c):max(c));
% make correlator
[L,num] = bwlabel(bw);
stats = regionprops(L);
rc = [];
for i = 1 : num
    bwi = imcrop(bw, round(stats(i).BoundingBox));
    bwi = imresize(bwi, size(bw2));
    rc(i) = corr2(bwi,bw2);
end
% max corr
[~, ind] = max(rc);
figure; imshow(img);
hold on; rectangle('Position', round(stats(ind).BoundingBox), 'EdgeColor', 'r', 'LineWidth', 2, 'LineStyle', '-');

[Rahul Shah]

Hello,

How can I generate plot to show the correlation peak that this highest intensity is for 'G'? I have attached an example of UTK where plot show the peak for 'T'.

[Image Analyst]

@Rahul Shah, youi can find the brightest point then get that line. If your correlation image is corrImage, do

maxValue = max(corrImage(:))
[rows, columns] = find(corrImage == maxValue);
verticalProfile = corrImage(:, columns(1));
plot(verticalProfile, 'r-', 'LineWidth', 2);
horizontalProfile = corrImage(rows(1), :);
plot(horizontalProfile, 'b-', 'LineWidth', 2);
ylabel('Correlation Value');
xlabel('Pixel Location')
title('Correlation Profile Through Peak')
grid on;
legend('Vertical', 'Horizontal')