My question is regarding classifying the images based on the whichever color is present more in the given image after converting the color image into gray and finding its histogram.

clc;
Image = imread('rose.jpg');
figure
imshow(Image);
I=rgb2gray(Image);
figure
imshow(I)
[count,x] = imhist(I)
index1 = find(x == 67);
index2 = find(x == 100);
numPixels = sum(count(index1:index2))

This code worked well up to finding the sum of number of pixels from one intensity range to other.

Now I am having few images. Some of them have red pixels in large amounts,some have white, some have blue, etc. I have to classify these images as blue, white, red, etc. after finding histogram and counting pixels in certain range. How can I do that?

Or is there any other way to do this classification?

Answer 1

[Image Analyst]

You just need to set up some rules, like if the mean red count from the histogram is more than the mean blue and green counts. See my File Exchange for color demos. http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862

[snehal jaipurkar]

sir i tried with some of codes given but i m not getting proper result. Can i send you two of the images which i have? one is more blue color and one is more in purple color. can u send me the code to classify the the purple image as to be shown "a purple image" and the blue image to be shown as " a blue image" in a dialogue box based on the count of pixels please? i used the following code to find the count of pixels using histogram.

clc;
Image = imread('water aura.jpg');
figure
imshow(Image);
I=rgb2gray(Image);
figure
imshow(I)
[count,x] = imhist(I)
index1 = find(x == 67);
index2 = find(x == 100);
numPixels = sum(count(index1:index2))
Image1 = imread('water aura 2.jpg');
figure
imshow(Image1);
s=rgb2gray(Image1);
figure
imshow(s)
[count,y] = imhist(s)
index3 = find(y == 67);
index4 = find(y == 100);
numPixels1 = sum(count(index3:index4))

[Image Analyst]

How do you expect to determine the color when the first thing you do upon reading in your color image is to convert it to grayscale????? You need to put some more thought into what you're doing and not do typical novice things like immediately calling rgb2gray() or edge() ( you're not calling edge but it seems like 90% of novices do for some reason that I can't figure out).

What you need to do do determine if the image is closer to purple or red is to convert each image to HSV color space and find the mean hue angle:

hsvImage = rgb2hsv(rgbImage);
hueImage = hsvImage(:,:,1);
sumsin = sum(sind(hueImage (:))) % Use (:) if you want to be able to handle 2-D hue images.
sumcos = sum(cosd(hueImage (:)))
averageHueAngle = atan2d(sumsin, sumcos) % Compute mean hue angle

Then you need to determine if the hue angle is closer to a purple hue angle or a red hue angle.

[snehal jaipurkar]

sir i tried the program which you gave:

hsvImage = rgb2hsv(water aura.jpg);
hueImage = hsvImage(:,:,1);
sumsin = sum(sind(hueImage (:))) % Use (:) if you want to be able to handle 2-D hue images.
sumcos = sum(cosd(hueImage (:)))
averageHueAngle = atan2d(sumsin, sumcos) % Compute mean hue angle

the following error is coming:

??? Error using ==> image
Error using ==> image
Numeric or logical matrix required for image CData

and the figure that is shown along with dis image is attached here. Sir I am very new to matlab, So i face problems in understanding.The images which i sent in previous question: water aura.jpg and water aura 2.jpg, in those images one kind of water has blue aura and the other has purple aura. i want to classify water samples based on the color of the aura each water sample has. can you suggest me the way i should choose for this classification of water samples based on their aura images???? In a paper i had read about such classification based on histogram. So i was trying that.

[Image Analyst]

You didn't try the program I gave you. You did this:

hsvImage = rgb2hsv(water aura.jpg);

while I had this:

    hsvImage = rgb2hsv(rgbImage);

You can't pass in a filename to rgb2hsv() and especially one without single quotes around it. You need to do this and read your image into a variable first.

filename = fullfile(pwd, 'water aura.jpg');
if exist(filename, 'file')
    rgbImage = imread(filename);
else
    % File does not exist
    message = sprintf('File does not exist: %s', filename);
    uiwait(errordlg(message));
    return;
end
hsvImage = rgb2hsv(rgbImage);

[snehal jaipurkar]

sir, i tried dis code:

filename = fullfile(pwd, 'water aura.jpg');
if exist(filename, 'file')
    rgbImage = imread(filename);
else
    % File does not exist
    message = sprintf('File does not exist: %s', filename);
    uiwait(errordlg(message));
    return;
end
hsvImage = rgb2hsv(rgbImage);

hueImage = hsvImage(:,:,1);
sumsin = sum(sind(hueImage (:))) % Use (:) if you want to be able to handle 2-D hue images.
sumcos = sum(cosd(hueImage (:)))
averageHueAngle = atan2d(sumsin, sumcos) % Compute mean hue angle

some error as such is shown:

sumsin =
  3.9034e+003
sumcos =
  3.1807e+005

??? Undefined function or method 'atan2d' for input arguments of type 'double'.

Error in ==> matlabgrp1 at 15
averageHueAngle = atan2d(sumsin, sumcos) % Compute mean hue angle

[Image Analyst]

What does this show:

>> which -all atan2d

and what release of MATLAB are you using? atan2d() is old - it was released way back in 2012 but maybe you have an ancient version of MATLAB.

[snehal jaipurkar]

??? Undefined function or method 'atan2d' for input arguments of type 'double'.

Error in ==> matlabgrp1 at 15 averageHueAngle = atan2d(sumsin, sumcos) % Compute mean hue angle

this error is shown sir for the code which is sent last time. and i am using 2010 version sir...and u had sent me this code: hsvImage = rgb2hsv(rgbImage); hueImage = hsvImage(:,:,1); sumsin = sum(sind(hueImage (:))) % Use (:) if you want to be able to handle 2-D hue images. sumcos = sum(cosd(hueImage (:))) averageHueAngle = atan2d(sumsin, sumcos) % Compute mean hue angle

i added dis only. but error as such (??? Undefined function or method 'atan2d' for input arguments of type 'double'. ) is shown sir...

[Image Analyst]

Did you see where I said " it was released way back in 2012 but maybe you have an ancient version of MATLAB." Well, 2010 is 2 years before atan2d was released. You'll have to see how you can build it yourself with atan() or atand(). Search the Mathworks web site for the description of atan2d().

[snehal jaipurkar]

Ok sir...I vl try to do that if I can get proper results by using atand( )... Thanks a lot for u help sir...