How to remove colour background of an image in MATLAB

Question

I want to remove the green pixels in this image and replace it with white background as a preliminary step to do canny detection on this picture to detect only the spanner. I converted it into hsv and considered h without green as follows, But didn't work. Please Help.

image = imread('F:\03.jpg');
hsv = rgb2hsv(image);
hChannel = hsv(:, :, 1);
sChannel = hsv(:, :, 2);
vChannel = hsv(:, :, 3);
newH = hsv(:,:,1) > 0.25 & hsv(:,:,1) < 0.41;
newV = (0.1) * vChannel;    % I am trying to change brightness
newHSVImage = cat(3, newH, sChannel, newV);
newRGBImage = hsv2rgb(newHSVImage);
imshow(newRGBIMage)

@SaurabhMistry Don't add superfluous code ticks to text. You added nothing useful to the post. When you're editing a question, make sure its value increases by your edit. I hope it was accidental and you were not just gaming the system for reputation — Sardar Usama, Jan 15 '18 at 12:04
next time use a smoother background with homoneneous saturation. that carpet is not very good — Piglet, Jan 15 '18 at 14:35
If you are able to remove the background then you've already detected the spanner. What's the point of applying Canny after that? — Cris Luengo, Jan 16 '18 at 03:17
I don't want to detect the spanner. I want to know about how perfect edge detection can be applied if there is a background which is not white. — Image Check, Jan 16 '18 at 09:12

ibezito · Accepted Answer · 2018-01-15T12:41:29.283

Solution

There are two main issues with your solution:

post-processing morphological operations are required, since some of the background pixels are not green (some of them are black).
it would be easier to add the white background on the rgb space.

Code

I suggest the following solution:

%generates mask of forground
fgMask = ~(hsv(:,:,1) > 0.25 & hsv(:,:,1) < 0.41);
CC = bwconncomp(fgMask);
numOfPixels = cellfun(@numel,CC.PixelIdxList);
[~,indexOfMax] = max(numOfPixels);
fgMask = zeros(size(fgMask));
fgMask(CC.PixelIdxList{indexOfMax}) = 1;

%morphological operations
fgMask = imopen(fgMask,strel('disk',2));
fgMask = imclose(fgMask,strel('disk',5));

%updating image in RGB space
rChannel = image(:, :, 1); rChannel(~fgMask) = 255;
gChannel = image(:, :, 2); gChannel(~fgMask) = 255;
bChannel = image(:, :, 3); bChannel(~fgMask) = 255;
image = cat(3, rChannel, gChannel, bChannel);

%display image
imshow(image)

Result

OP wants to apply Canny to the result. Note that, given `fgMask` as computed here, `imdilate(fgMask,strel('diamond',1))-fgMask` gives the edges of the object. Applying Canny would be superfluous. — Cris Luengo, Jan 16 '18 at 07:19

score 3 · Answer 2 · answered Jan 15 '18 at 12:15

You don't seem to understand what you are doing. With comments:

% Select only green indexes
newH = hsv(:,:,1) > 0.25 & hsv(:,:,1) < 0.41;

% Change brigthness of the whole image
newV = (0.1) * vChannel;

What your code does is, gets the logical index of all green pixels, and reduces brigthness of the whole image. Then, you use the logical indexes as color value, thus, if you do newV = (1) * vChannel; and plot, you will realize that all your green is now red (red: HSV=1).

What you want is select the green, and reduce the brightness (or whatever you want to do) of specifically the green.

for that, do:

% Select only green indexes
green_index = hsv(:,:,1) > 0.25 & hsv(:,:,1) < 0.41;
% change the brigtness of those specific pixels
newV=vChannel;
newV(green_index)=0.1*newV(green_index);

newHSVImage = cat(3, hChannel, sChannel, newV);

You may need to tune the range of your green detection in H.

Thanks @AnderBiguri That was right and I got my green pixels to be red. I now understand why it happened. What I was trying to do is replace the green pixels to white, So that I can have a clear background to do edge detection. I am just changing the brightness value of the image to see how edge is influenced. I will study more on it. Thanks. — Image Check, Jan 15 '18 at 20:35

How to remove colour background of an image in MATLAB

2 Answers2