Thresholding an image based on part of a histogram's values

Hi folks,

I'm trying to get the parts of an image that fall between 2 values of its grayscale histogram, the revert the image back to its colour form. So far, I am struggling to get anything but a black image! Any ideas on what I'm doing wrong please? Below is my code.

baseGray = rgb2gray(I);

isoLower = 135;
isoUpper = 155;

lessThan = baseGray < isoLower;
greaterThan = baseGray > isoUpper;

baseGray(lessThan) = 0;
baseGray(greaterThan) = 0;
grayThresh = imbinarize(baseGray);
grayThresh = bwareaopen(grayThresh, 3000);
grayThresh = imfill(grayThresh, 'holes');
grayThresh = bwperim(grayThresh);
grayThresh = imdilate(grayThresh, ones(5));
grayThresh = imerode(grayThresh, ones(3));
grayThresh = imfill(grayThresh, 'holes');

refigure = img.*repmat(uint8(grayThresh),[1 1 3]);
imshow(refigure);

Answer 1

[Ameer Hamza]

Try something like this

img = im2double(imread('pears.png'));
img_gray = rgb2gray(img);

isoLower = 135/255; % double image, pixel intensity between 0 and 1
isoUpper = 155/255;

mask = (isoLower < img_gray) & (img_gray < isoUpper);
refigure = img;
refigure(~mask(:,:,[1 1 1])) = 0;
imshow(refigure)

[Teshan Rezel]

Hi Ameer, I tried this but I only get the same image as the input as an output...any idea what might be wrong? In other words, the segmentation part doesn't seem to work...

baseGray = rgb2gray(I);

isoLower = 135/255;
isoUpper = 155/255;

mask = (baseGray >= isoLower) & (baseGray <= isoUpper);
I(~mask(:,:,[1 1 1])) = 0;
imshow(I);

[Ameer Hamza]

Can you show how image 'I' is created? Also, can you show the output of

class(I)

before running the lines of code in your comment?

[Teshan Rezel]

Hi Ameer,

Sure thing. The whole code is as follows:

img = imread('25_04_2018_102541.JPG');

imgGrey = rgb2gray(img);
[counts, ~] = imhist(imgGrey, 255);
T = otsuthresh(counts);
BW = imbinarize(imgGrey, T);
BW = bwareaopen(BW, 3000);
BW = imfill(BW, 'holes');
BW = bwperim(BW);
BW = imdilate(BW, ones(5));
BW = imerode(BW, ones(3));
BW = imfill(BW, 'holes');
I = img.*repmat(uint8(BW),[1 1 3]);

baseGray = rgb2gray(I);

isoLower = 135/255;
isoUpper = 155/255;

mask = (baseGray > isoLower) & (baseGray < isoUpper);
I(~mask(:,:,[1 1 1])) = 0;
imshow(I);

The outputs for class img and I are as follows (respectively):

[Ameer Hamza]

Sine your image is in uint8 format, the pixel values vary from 0 to 255. Therefore, change the limits too

isoLower = 135;
isoUpper = 155;

[Teshan Rezel]

Thanks Ameer! However, I get something that looks rather strange...would you happen to know how to fix this please?

[Ameer Hamza]

I guess this is the expected result since you set the limits set the grayscale limits in a very narrow range (135 to 155). You need to change the range to get a different result.

[Teshan Rezel]

That makes sense...thank you for your help!

[Ameer Hamza]

I am glad to be of help!

[Image Analyst]

Why not just try the Color Thresholder on the Apps tab of the tool ribbon and interactively set your thresholds???

[Teshan Rezel]

Hi Image Analyst, I've tried the tool and found it useful to a degree. But my issue is that I will need to analyse several batches of 900 images each, so doing them individually in the app would not be efficient. Furthermore, I need to analyse them in grayscale, which isn't an option on the app as far as I can see. Thanks for your suggestion!