Problem in plotting the matrix over an image.

I am required to read an image,find out its rgb channels and then in those channels separately,mark the pixels whose value exceeds above a certain threshold value.I wrote the following code for marking the pixels but they are not getting marked and I am getting the error that it exceeds matrix dimensions.What am I doing wrong in the code? Below is the code-

img=imread('test.jpg');

imshow(img)

redchannel=img(:,:,1);

greenchannel=img(:,:,2);

bluechannel=img(:,:,3);

rs=zeros(size(redchannel));

r1=cat(3,redchannel,rs,rs);

figure

imshow(r1)

gs=zeros(size(greenchannel));

g1=cat(3,gs,greenchannel,gs);

figure

imshow(g1)

bs=zeros(size(bluechannel));

b1=cat(3,bs,bs,bluechannel);

figure

imshow(b1)

[r,c,~]=size(img);

indx=zeros(r,c);

for i=1:r

    for j=1:c

        if ((redchannel(i,j))>=180)

            indx(i,j)=1;

        end

    end

end

for i=1:r

    for j=1:c

imshow(r1)

hold on

plot(indx(:,1),indx(:,2),'b*');

hold off

    end

end

[KSSV]

Your code is taking hell lot of time: Use below modified code.

img=imread('Prachi Sharma image');
imshow(img)
%
redchannel=img(:,:,1);
greenchannel=img(:,:,2);
bluechannel=img(:,:,3);
%
rs=zeros(size(redchannel));
r1=cat(3,redchannel,rs,rs);
figure
imshow(r1)
%
gs=zeros(size(greenchannel));
g1=cat(3,gs,greenchannel,gs);
figure
imshow(g1)
%
bs=zeros(size(bluechannel));
b1=cat(3,bs,bs,bluechannel);
figure
imshow(b1)
%
[r,c,~]=size(img);
indx=zeros(r,c);
indx(redchannel>=180) = 1 ;
% for i=1:r
%     for j=1:c
%         if ((redchannel(i,j))>=180)
%             indx(i,j)=1;
%         end
%     end
% end
for i=1:r
    for j=1:c
        imshow(r1)
        hold on
        plot(indx(:,1),indx(:,2),'b*');
        drawnow
        hold off
    end
end

Can you tell me what you are trying to do here? Why loop needed here?

for i=1:r
      for j=1:c
          imshow(r1)
          hold on
          plot(indx(:,1),indx(:,2),'b*');
          drawnow
          hold off
      end
  end

[Prachi Sharma]

I am trying to check the matrix of every channel and when the value of any pixel in the channel exceeds a certain threshold value(in this case 180) it should get highlighted

[Guillaume]

Important note:

redchannel = img(:,:,1);
rs = zeros(size(redchannel));
r1 = cat(3, redchannel, rs, rs);

That's very dangerous because there's no class specified in the zeros call. rs is always going to be of class double, whereas redchannel may be of a different class (if it comes from a jpeg image, it'll be uint8). You're then combining matrices of different classes which may not produce the result you expect.

Much safer would be:

redchannel = img(:,:,1);
rs = zeros(size(redchannel), class(redchannel));  %make sure rs is of the same class as redchannel
r1 = cat(3, redchannel, rs, rs);

or even better:

r1 = img;  %r1 is same class as img
r1(:, :, [2 3]) = 0; %note that the double 0 gets converted to 0 of the class of r1

See this question and this question for how easy it is to be bitten by this.

[Image Analyst]

Do you want to change the actual pixel values, like make them red or something? OR do you want to leave the pixel values the same and just change how they are displayed, like use a colormap where those pixels are red, or use a binary overlay ?

[Prachi Sharma]

I need to take an image,find its 3 channels and then in every channel I need to check if any pixel is exceeding a certain threshold value.If a pixel exceeds a certain threshold value then those pixels should be marked or highlighted in the image of that channel.

Answer 1

[KSSV]

clc; clear all ;
img=imread('Prachi Sharma image');
imshow(img)
%
redchannel=img(:,:,1);
greenchannel=img(:,:,2);
bluechannel=img(:,:,3);
%
rs=zeros(size(redchannel));
r1=cat(3,redchannel,rs,rs);
figure
imshow(r1)
%
gs=zeros(size(greenchannel));
g1=cat(3,gs,greenchannel,gs);
figure
imshow(g1)
%
bs=zeros(size(bluechannel));
b1=cat(3,bs,bs,bluechannel);
figure
imshow(b1)
%
[r,c,~]=size(img);
indx=NaN(r,c);
indx(redchannel>=180) = 1 ;
%
imshow(r1) ;
x = 1:c ;
y = 1:r ;
[X,Y] = meshgrid(x,y) ;
hold on
plot3(X,Y,indx,'*b') ;

[Prachi Sharma]

Hey thank you it is working.I do have a question though,why have you taken first loop till columns and another loop till rows.Why can't it be the other way round?

[KSSV]

You try it out...this is the way the matrix size of x,y making equal to indx. Do consider the Guillaume comments.

[Prachi Sharma]

I tried it out,switching places of r and c and it didn't work that way.

Answer 2

[Guillaume]

find will return the locations of all the values above the threshold, there's no need for a loop

img = imread('test.jpg');
redchannel = img(:, :, 1);
[row, column] = find(redchannel > 180);  %no need for loop
redimage = img; redimage(:, :, [2 3]) = 0;
figure;
imshow(redimage);
hold on;
plot(column, row, 'b*');  %column, row <=> x, y. Stupid matlab can't decide on a single coordinate system!

and do the same for green and blue.

edit: also see my comment to the question about using zeros with no class specified.

[Prachi Sharma]

Hey thanks it is working.I do have one more question,can you please explain this command "redimage(:,:,[2,3])=0",I could not understand [2,3] part in this. Also in the last command plot(column,row,'b*') why are you taking columns first and not rows.

P.S-The above question might sound stupid but I need to learn and not just copy.Thanks.

[Guillaume]

redimage(:, :, [2 3]) selects all the rows, all the columns and only index 2 and 3 of the third dimension. index 2 is the green channel, and index 3 is the blue channel. In effect, this set all the green and blue value to 0. If you wanted to keep just the green channel, then:

img(:, :, [1 3]) = 0; %1and 3 is red and blue respectively

find use the normal matlab convention where the first value it returns is the row of the matrix, and the second value is the column of the matrix. plot and a few other functions (most of of the image processing toolbox, as well as gradient, meshgrid and maybe more) use a different coordinate system (x, y) where x corresponds to the columns, and y to the rows. This is matches the way you normally references coordinates outside matlab but as a result you end up with two different coordinates system where one is swapped compared to the other.

Personally, I think it's stupid (hence my comment in the code). In my opinion it would be much better to be internally consistent but that's what it is: plot uses (x, y) coordinates which is (column, row), find uses (row, column) which is (y, x).

[Image Analyst]

MATLAB is not the only one. Have you ever noticed the two conventions for numeric keypads - how your keyboard and phone are upside down with respect to each other? But yes, x,y and row,column trips up lots of people.

[Prachi Sharma]

@guillaume and @image analyst..Thanks a lot guys.

[Prachi Sharma]

@guillaume..Your code is not making the pixels who exceeds a certain threshold value zero in every channel.

[Guillaume]

No, it doesn't and it can't. The code I've written does exactly and only this:

• find the locations of pixels whose red pixels are above the threshold
• creates a duplicate of the image where only the red channel is active
• displays the image
• overlays on the display blue stars at the locations found in the 1st step.

At no point does it alter the value of pixels above any threshold.

[Prachi Sharma]

Ok that works for me.I only have one concern left though.I need to "logical or" all the locations where the pixel values goes above the threshold in all the three channels. How do I do that?

[KSSV]

As suggested by Guillaume in other discussion.

[Y,X,V] = redchannel>=180 ;

Follow the same for other channels.

[Prachi Sharma]

I need to have different thresholds for every channel.

[KSSV]

Dear friend then type it..what is stopping you?

[Prachi Sharma]

This will also work right?

%for redchannel

mask_r1=(redchannel<x);

newm1=uint8(mask_r1);

mult1=newm1.*redchannel;

%for greenchannel

mask_r2=(greenchannel<y);

newm2=uint8(mask_r2);

mult2=newm2.*greenchannel;

%for blue channel

mask_r3=(bluechannel<z);

newm3=uint8(mask_r3);

mult3=newm3.*bluechannel;

%or ing the result

ora=(newm1|newm1)|newm3;

uora=uint8(ora);

figure

imagesc(uora)

title('Ored result of all the threshold values');

Answer 3

[Image Analyst]

Try something like this:

% Extract the individual red, green, and blue color channels.
redChannel = rgbImage(:, :, 1);
greenChannel = rgbImage(:, :, 2);
blueChannel = rgbImage(:, :, 3);
% Make mask where ANY of the color channels is saturated.
saturatedPixels = redChannel  == 255 | greenChannel  == 255 | blueChannel == 255;
% Make red channel 255, and other color channels 0 where there is saturation
redChannel(saturatedPixels) = 255;
greenChannel(saturatedPixels) = 0;
blueChannel(saturatedPixels) = 0;
% Recombine separate color channels into a single, true color RGB image.
rgbImage = cat(3, redChannel, greenChannel, blueChannel);
imshow(rgbImage);

Now your RGB image will be pure red wherever ANY pixel of ANY color channel is saturated (value of 255). For an image in the range 0-1, use 1 instead of 255. For a uint16 image, use 65535 instead of 255.