How can I know the value of pixel "ROI" in each (RGB) Channel?

I hope someone can help me out.

I have 4 ROI Buildings, Cars, Street and Trees and each one has defined but what i need to know, How can i get the Values of pixels of ROI in each channel in RGB image (Red, Green and Blue)?

Please tell me exactly, the code, that i should write!

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

% CleanUp
clc;
clear;
close all;

%Read Photo RGB and DSM
orph = imread('BE_ORTHO_27032011_315145_56865.tif');
figure, imshow (orph,[]);

%Define Region of interest
roi_Building = 	drawpolygon('Color','r');
roi_Street = drawline('Color','y');
roi_Tree = drawfreehand('Color','g');
roi_car = 	drawpolygon('Color','b');

%Seperate channels RGB
op_red= orph(:,:,1);   % Red channel
op_green= orph(:,:,2); % Green channel
op_blue= orph(:,:,3);  % Blue channel

%Value of pixels in all layers

Answer 1

[Walter Roberson]

createMask

[r, c, p] = size(orph);
mask_building = createMask(roi_Building, r, c);

building_pixels = [op_red(mask_building), op_green(mask_building), op_blue(mask_building)];

This would be N x 3 where N is the number of selected pixels, and the columns are R, G, B.

But I wonder if what you would prefer would instead be

[r, c, p] = size(orph);
mask3_building = repmat(createMask(roi_Building, r, c), [1, 1, 3]);
masked_building = zeros(r, c, p, 'like', orph);
masked_building(mask3_building) = orph(mask3_building);

which would create a new imae masked_building in which the pixels that are not inside the ROI are all set to 0 (black)

[Amr Azab]

Thank you So much ! :)

[Amr Azab]

Hello Mr Walter @Walter Roberson

i hope that you help me out.

The Matlab gives me no results with this "value masked_building(mask3_building)"

The Average (AVG_BU ) is just 1*1 it should be 1*1*3 How can i do that ? (See photo)

what is wrong in my code?

------------------------------------------------------------------------------------------------------------------

% CleanUp

clc;

clear;

close all;

%Read Photo RGB and DSM

orph = imread('BE_ORTHO_27032011_315145_56865.tif');

figure, imshow (orph,[]);

%Define Region of interest

roi_Building = drawpolygon('Color','r');

roi_ground = drawline('Color','y');

roi_vegetation = drawfreehand('Color','g');

%Seperate channels RGB

op_red = orph(:,:,1); % Red channel

op_green = orph(:,:,2); % Green channel

op_blue = orph(:,:,3); % Blue channel

%Value of pixels for ROI in all layers

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

mask_building = createMask(roi_Building, r, c);

mask_ground = createMask(roi_ground, r, c);

mask_vegetation = createMask(roi_vegetation, r, c);

building_pixels = [op_red(mask_building), op_green(mask_building), op_blue(mask_building)];

ground_pixels = [op_red(mask_ground), op_green(mask_ground), op_blue(mask_ground)];

vegetation_pixels = [op_red(mask_vegetation), op_green(mask_vegetation), op_blue(mask_vegetation)];

%Buildings

[r, c, p] = size(orph); %Row, Column and layers (size of Image)

mask3_building = repmat(createMask(roi_Building, r, c), [1, 1, 3]); %Repeat copies of array

masked_building = zeros(r, c, p, 'like', orph);

masked_building(mask3_building) = orph(mask3_building);

%Ground

mask3_ground = repmat(createMask(roi_ground, r, c), [1, 1, 3]);

masked_ground = zeros(r, c, p, 'like', orph);

masked_ground(mask3_ground) = orph(mask3_ground);

%Vegitation

mask3_vegetation = repmat(createMask(roi_vegetation, r, c), [1, 1, 3]);

masked_vegetation = zeros(r, c, p, 'like', orph);

masked_vegetation(mask3_vegetation) = orph(mask3_vegetation);

%Arithmetic average for each Class

AVG_BU = mean(masked_building(mask3_building));

AVG_GR = mean(masked_ground(mask3_ground));

AVG_VG = mean(masked_vegetation(mask3_vegetation));

%Distance between every Pixel in Picture to average for each Class

i=1:10000;

j=1:10000;

or = [i,j];

Class_mean1 = ;

Class_mean2 = ;

Class_mean3 = ;

Di_Building = sqrt((or(:,:,1)-Class_mean1(1,1)).^2 + (or(:,:,2)-Class_mean1(2,1)).^2 + (or(:,:,3)-Class_mean1(3,1)).^2);

Di_ground = sqrt((or(:,:,1)-Class_mean2(1,1)).^2 + (or(:,:,2)-Class_mean2(2,1)).^2 + (or(:,:,3)-Class_mean2(3,1)).^2);

Di_Vegetation = sqrt((or(:,:,1)-Class_mean3(1,1)).^2 + (or(:,:,2)-Class_mean3(2,1)).^2 + (or(:,:,3)-Class_mean3(3,1)).^2);

[Walter Roberson]

You have an irregular roi, something that is not a rectangle . You want to extract the pixels from inside the roi. The result cannot be rectangular because the roi is not rectangular . You are using logical indexing to get the pixel component values within the roi .The result of logical indexing is always a column vector . Mean of a vector is always a scalar .

If you need the mean for each color plane then you have three options:

• loop over the image locations. If the pixel is inside the roi then add the components to the bottom of an Nx3 array
• use logical indexing on each color pane, getting a column vector for each, and putting the vectors together into three columns
• provided that you repmat the mask into the third dimension, use logical indexing of the whole matrix and then reshape the results to have three columns

[Amr Azab]

Thank you so much. What i need is, the mean for each Class (Buildings, Ground and Vegetation) in all Layers of Image (R,G and B)

Sould all ROI (rectangular) as you mentioned?

When you have time, please give me an example of codes that i should do? because i am not good enough in coding. I am still learning.

[Walter Roberson]

You already did the work of extracting the pixels in

building_pixels = [op_red(mask_building), op_green(mask_building), op_blue(mask_building)];
ground_pixels = [op_red(mask_ground), op_green(mask_ground), op_blue(mask_ground)];
vegetation_pixels = [op_red(mask_vegetation), op_green(mask_vegetation), op_blue(mask_vegetation)];

You can mean() each of those instead of working with masked_building and so on.

masked_building and so on exist for the case where you need a copy of the image (full size) in which everything except the building (or ground or vegetation) has been set to black, such as if you wanted to display the masked items in context.

[Amr Azab]

Sorry Mr Walter

I couldnt get 1*1*3 until now

Which part of code should be change?

and I got always Error with Di_Building, ground and Vegetation

----------------------------------------------------------------------------------------------------------------

% CleanUp

clc;

clear;

close all;

%Read Photo RGB and DSM

orph = imread('BE_ORTHO_27032011_315145_56865.tif');

figure, imshow (orph,[]);

%Define Region of interest

roi_Building = drawrectangle('Color','r');

roi_ground = drawrectangle('Color','y');

roi_vegetation = drawrectangle('Color','g');

%Seperate channels RGB

op_red = orph(:,:,1); % Red channel

op_green = orph(:,:,2); % Green channel

op_blue = orph(:,:,3); % Blue channel

%DSM photo

%DSM = imread('BE_DSM_27032011_315145_56865.tif');

%figure, imshow (dsm,[]), title ('DSM');

%Value of pixels for ROI in all layers

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

mask_building = createMask(roi_Building, r, c);

mask_ground = createMask(roi_ground, r, c);

mask_vegetation = createMask(roi_vegetation, r, c);

building_pixels = [op_red(mask_building), op_green(mask_building), op_blue(mask_building)];

ground_pixels = [op_red(mask_ground), op_green(mask_ground), op_blue(mask_ground)];

vegetation_pixels = [op_red(mask_vegetation), op_green(mask_vegetation), op_blue(mask_vegetation)];

%Arithmetic average for each Class

AVG_BU = mean(building_pixels);

AVG_GR = mean(ground_pixels);

AVG_VG = mean(vegetation_pixels);

Class_mean1 = size(AVG_BU);

Class_mean2 = size(AVG_GR);

Class_mean3 = size(AVG_VG);

Di_Building = sqrt((orph(:,:,1)-Class_mean1(1,1)).^2 + (orph(:,:,2)-Class_mean1(2,1)).^2 + (orph(:,:,3)-Class_mean1(3,1)).^2);

Di_ground = sqrt((orph(:,:,1)-Class_mean2(1,1)).^2 + (orph(:,:,2)-Class_mean2(2,1)).^2 + (orph(:,:,3)-Class_mean2(3,1)).^2);

Di_Vegetation = sqrt((orph(:,:,1)-Class_mean3(1,1)).^2 + (orph(:,:,2)-Class_mean3(2,1)).^2 + (orph(:,:,3)-Class_mean3(3,1)).^2);

[Walter Roberson]

Class_mean1 = size(AVG_BU);

You are taking the size() of the averages. Notice that in each case the size is 1 x 3, so the Class variables are each the two element vectors [1, 3]

Di_Building   =  sqrt((orph(:,:,1)-Class_mean1(1,1)).^2 + (orph(:,:,2)-Class_mean1(2,1)).^2 + (orph(:,:,3)-Class_mean1(3,1)).^2);

You index the 1 x 2 vector of size information at up to index (3,1) but the row vector of size information does not have 3 rows.

I suggest that you want to be subtracting AVG_BU(1) AVG_BU(2) AVG_BU(3)

[Amr Azab]

Thank you Mr Walter