Quantization HSV Space by assigning 8 level each to hue, saturation and value to give a quantized HSV space with 8x8x8=512 histogram bins

Hello,

I will try to quantize the HSV color space but I can not, can anyone help me please?

I want to Quantization HSV Space by assigning 8 level each to hue, saturation and value to give a quantized HSV space with 8x8x8=512 histogram bins.

Thanks

Answer 1

[Subhadeep Koley]

Here is an example of quantization of the HSV colorspace

clc; close all;
rgbImage = imread('peppers.png');
hsvImage = rgb2hsv(rgbImage);
montage(hsvImage,'Size', [1, 3]); title('Before Quantization'); colorbar;
threshRGB = multithresh(hsvImage, 8);
threshForPlanes = zeros(3, 8);
for i = 1:3
threshForPlanes(i, :) = multithresh(hsvImage(:, :, i), 8);
end
quantPlane = zeros(size(hsvImage));
for i = 1:3
    value = [0 threshForPlanes(i, 2:end) 1]; 
    quantPlane(:, :, i) = imquantize(hsvImage(:, :, i), threshForPlanes(i, :), value);
end
quantPlane = double(rescale(quantPlane, 0, 1));
figure; montage(quantPlane, 'Size', [1,3]); title('After Quantization'); 
colormap(parula(8)); colorbar;

Hope this helps!

[shalaw faraj]

Dear Subhadeep,

Thanks for your help, but please I have one question, do you think the number 0, 255 is right for HSV color space as same of RGB?

[Subhadeep Koley]

@ shalaw faraj Yes, you are right, [0 - 1] range is better for HSV.

I have edited the answer. Thanks!

[shalaw faraj]

Dear Subhadeep,

Do you think my output was right?

I think no difference between the after and before.

[Subhadeep Koley]

I think you're displaying the same image twice. Can you share the code which you've used to generate the above figures?

[shalaw faraj]

clc; close all;
rgbImage = imread('peppers.png');
hsvImage = rgb2hsv(rgbImage);
montage(hsvImage,'Size', [1, 3]); title('Before Quantization'); colorbar;
threshRGB = multithresh(hsvImage, 8);
threshForPlanes = zeros(3, 8);
for i = 1:3
threshForPlanes(i, :) = multithresh(hsvImage(:, :, i), 8);
end
quantPlane = zeros(size(hsvImage));
for i = 1:3
    value = [0 threshForPlanes(i, 2:end) 1]; 
    quantPlane(:, :, i) = imquantize(hsvImage(:, :, i), threshForPlanes(i, :), value);
end
quantPlane = double(rescale(quantPlane, 0, 1));
figure; montage(quantPlane, 'Size', [1,3]); title('After Quantization'); 
colormap(parula(8)); colorbar;

[Subhadeep Koley]

@ shalaw faraj I am getting the correct output when executing the above code.

Which MATLAB version are you using? Also you can try to convert image to double precision before rgb2hsv like below.

rgbImage = im2double(imread('peppers.png'));

[shalaw faraj]

Matlab R2017b

but why we have just one subplot and the other two are black? also you have three HSV image output.

I tried your new code im2double but it was the same.

[Subhadeep Koley]

@ shalaw faraj I have been able to reproduce the same issue in R2017b, earlier I was trying it in R2019b.

It seems it is a problem of the function montage, which is not scaling the images properly (this issue is addressed in later releases). However you can use subplot instead of montage. Refer the code below.

Hope this helps!

clc; close all;
rgbImage = im2double(imread('peppers.png'));
hsvImage = rgb2hsv(rgbImage);
figure('Name', 'Before Quantization',...
    'units','normalized','outerposition',[0 0 1 1]);
subplot(1, 3, 1); imshow(hsvImage(:, :, 1)); colorbar;
subplot(1, 3, 2); imshow(hsvImage(:, :, 2)); colorbar;
subplot(1, 3, 3); imshow(hsvImage(:, :, 3)); colorbar;
threshRGB = multithresh(hsvImage, 8);
threshForPlanes = zeros(3, 8);
for i = 1:3
    threshForPlanes(i, :) = multithresh(hsvImage(:, :, i), 8);
end
quantPlane = zeros(size(hsvImage));
for i = 1:3
    value = [0 threshForPlanes(i, 2:end) 1];
    quantPlane(:, :, i) = imquantize(hsvImage(:, :, i), threshForPlanes(i, :), value);
end
quantPlane = double(rescale(quantPlane, 0, 1));
figure('Name', 'After Quantization',...
    'units','normalized','outerposition',[0 0 1 1]);
subplot(1, 3, 1); imshow(quantPlane(:, :, 1)); colormap(gca, parula(8)); colorbar;
subplot(1, 3, 2); imshow(quantPlane(:, :, 2)); colormap(gca, parula(8)); colorbar;
subplot(1, 3, 3); imshow(quantPlane(:, :, 3)); colormap(gca, parula(8)); colorbar;

[shalaw faraj]

It was very helpful, I don't know how can I thanks to you, I have the last question please

how can I equalization separately, for example (18,3,3) instead of (8,8,8).

[Subhadeep Koley]

@ shalaw faraj Instead of using loop you can quantize them individually according to your need

[shalaw faraj]

I got it thanks again.

[shalaw faraj]

Dear @Subhadeep,

According my question I want to quantize the image RGB from 256*256*256 to 4*4*4 which it give 64 bin of histogram can you help me please?

thanks

Answer 2

[Fowzi barznji]

Sir

Subhadeep Koley

which version of MATLAB u r using ?