Remove white space from around image

Hello, I'm pretty new to MatLab, so I'm having a hard time doing some things with it. I have a grayscale jpg-image, with a very large white space around the actual image, which is dark. Is there an easy command, to remove the colors <250 from the image?

Regards, Martin

Answer 1

[DGM]

It's never too late for reference answers ...

Reading the question, I assume that the intent is to crop the excess padding from the image, rather than to merely change it to black. I also assume that it's desired to do this automatically, or otherwise without the need to tediously calculate the RECT parameter for a call to imcrop() or direct subscript indexing.

There are a few tools on the File Exchange that can do this. @Reza Farrahi Moghaddam posted one which (as the name implies) can automatically crop off white padding. There is also this tool designed for cropping padding of uniform intensity. Lastly, MIMT has a function called cropborder() which can crop manually or automatically based on user-selected metrics. Let's demonstrate all three.

First, let's generate some test images. In each case, the image is displayed with a narrow black border simply so that it shows up against the awful white background of this webpage. The black border isn't part of the working image.

% original image
A = imread('peppers.png'); % 384x512
imshow(A);

% image with uniform-width solid white matting
B = padarray(A,[10 10],255,'both');
imshow(padarray(B,[2 2]));

% image with nonuniform-width solid white matting
C = padarray(A,[5 7],255,'pre');
C = padarray(C,[15 13],255,'post');
imshow(padarray(C,[2 2]));

% image with uniform-width solid gray matting
D = padarray(A,[10 10],200,'both');
imshow(padarray(D,[2 2]));

% image with noisy white padding
s = size(A);
E = uint8(randi([245 256],s(1)+20,s(2)+20,size(A,3)));
E(11:s(1)+10,11:s(2)+10,:) = A;
imshow(padarray(E,[2 2]));

If each of these images is passed to the aforementioned cropping tools, we can see what works and what doesn't. Again, the name implies that RemoveWhiteSpace() works only on white padding. It works well, but only if the padding is solid white.

% use RemoveWhiteSpace()
S1(1,:) = size(RemoveWhiteSpace(B)); % works fine for solid white padding
S1(2,:) = size(RemoveWhiteSpace(C)); % even uneven padding
S1(3,:) = size(RemoveWhiteSpace(D)); % but not unless it's white
S1(4,:) = size(RemoveWhiteSpace(E)); % and not unless it's solid
S1

S1 =

384 512 3

384 512 3

404 532 3

404 532 3

As its description suggests, crop_img() crops off padding of any solid color.

% use crop_img()
S2(1,:) = size(crop_img(B)); % works fine for uniform padding
S2(2,:) = size(crop_img(C)); % or nonuniform padding
S2(3,:) = size(crop_img(D)); % of any solid color
S2(4,:) = size(crop_img(E)); % but not unless it's solid
S2

S2 =

384 512 3

384 512 3

384 512 3

404 532 3

On the other hand, cropborder() looks for a change in edge variance when it searches for the image extents. It can crop noisy or patterned borders.

% use cropborder()
S3(1,:) = size(cropborder(B)); % works fine for uniform padding
S3(2,:) = size(cropborder(C,[NaN NaN NaN NaN])); % or nonuniform padding
S3(3,:) = size(cropborder(D)); % of any solid color
S3(4,:) = size(cropborder(E)); % or even patterned or noisy color
S3

S3 =

384 512 3

384 512 3

384 512 3

384 512 3

Answer 2

[Sean de Wolski]

I(I>250) = 0;  %sets them to black.

To to crop them you'll have to use a cropping tool such as imcrop or know the indices you wish to crop to.

Answer 3

[Reza Farrahi Moghaddam]

You can try this tiny tool to remove white space around images:

http://www.mathworks.com/matlabcentral/fileexchange/34898-remove-white-space-around-images

Answer 4

[hf fh]

I used the same method as link http://www.mathworks.com/matlabcentral/fileexchange/34898-remove-white-space-around-images

But the problem is I have 20 images and I want to remove white space around images: How to apply 20 images to you the way I try but I can not..

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%dir folder 
input_filename = '/Users/mac/Desktop/RemoveWhiteSpace/RemoveWhiteSpace/';

% % Check to make sure that folder actually exists.  Warn user if it doesn't.

if ~isdir(input_filename)

    errorMessage = sprintf('Error: The following folder does not exist:\n%s', input_filename);

    uiwait(warndlg(errorMessage));

    return;
  end

% Get a list of all files in the folder with the desired file name pattern.

filePattern = fullfile(input_filename, '*.png'); % Change to whatever pattern you need.

theFiles = dir(filePattern);

for k = 1 :20

    baseFileName = theFiles(k).name;

    fullFileName = fullfile(input_filename, baseFileName);

    u_in = mat2gray(imread(fullFileName));

   imshow(u_in);

figure(k), 

imshow(RemoveWhiteSpace(u_in));

RemoveWhiteSpace([], 'file', '*.png', 'output', 'test_out*.png');

saveas(figure(k),sprintf('%d.tif',k))

end

Answer 5

[hf fh]

this good for 100 images

folder = '/Users/mac/Desktop/fiel2/RemoveWhiteSpace/RemoveWhiteSpace/semple';

%image3d = zeros(300, 300, 99, 'uint8');

for k = 1 : 99

         % Read one image.
       u_in = mat2gray(imread( fullfile(folder, sprintf('%d.tif', k) )));

        %figure, imshow(u_in);

       figure(k), imshow(RemoveWhiteSpace(u_in));

RemoveWhiteSpace([], 'file', sprintf('%d.tif', k) , 'output', sprintf('%d.png', k));

end

Answer 6

[Image Analyst]

To remove a frame/border from a gray scale image you can simply use any() to determine which rows or columns are not part of the padding and extract only those rows:

grayImage = grayImage(any(grayImage~=paddingValue, 2), any(grayImage~=paddingValue, 1));

Here is a full demo with comment, displays, etc.:

% Demo by Image Analyst.
% Crop off uniform padding from outside border of image.
% This code is general enough to work with any padding value: 0, 255, or anything else.
clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
clearvars;
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 16;
% Create the image padded with values of 162 all around. (Could be 255 or any value actually).
grayImage = imread('cameraman.tif');
% Pad the image
grayImage = padarray(grayImage, [15, 70], 162);
subplot(2, 2, 1)
imshow(grayImage);
impixelinfo
axis('on', 'image')
title('Original Padded Image')

% Determine the value of the padding by looking at the upper left pixel.
paddingGrayLevel = grayImage(1,1)
% Get a mask showing all pixels where the value is the padding value.
mask = grayImage == paddingGrayLevel;
subplot(2, 2, 2)
imshow(mask);
axis('on', 'image')
title('Mask of Padded Values')

% Find rows and columns where any of the values are NOT the padding gray level.
% If at least one value is not the padding value,
% then the row or column is not part of the padding/frame.
goodRows = any(~mask, 2);
goodColumns = any(~mask, 1);
% Extract the good rows and columns.
grayImage = grayImage(goodRows, goodColumns);
subplot(2, 2, 3)
imshow(grayImage);
axis('on', 'image')
title('Cropped Image with No Padding')

[DGM]

That's pretty compact for being as direct as it is.

Consider the counterexamples:

I could see the case where this is desired, but I don't think it would generally be the intent.

A = imread('testmontage.png');
imshow(A)

paddingValue = 128;
B = A(any(A~=paddingValue, 2), any(A~=paddingValue, 1));
clf; imshow(B)

While it might seem like nonuniform padding is something that nobody would ever add, pretty much any JPG will have a ring of crust around it. Answering things on the forum has taught me that this is pretty ubiquitous, considering MATLAB's low-fidelity defaults when saving images and figures as JPG, and the routine need to strip padding off of images saved from figures.

A = imread('noisytest.jpg');
clf; imshow(A)

paddingValue = 128;
B = A(any(A~=paddingValue, 2), any(A~=paddingValue, 1));
clf; imshow(B)

I suppose some tolerance might help with that.

[Image Analyst]

My code was considering a single photo, not a group of them. So for a group of them it will just give the overall bounding box of the whole group. It will not split them apart into separate images and remove padding on each image.

Hopefully everyone knows not to save images as JPG but we all know that is not true and still some people still use JPG. You are correct that JPG can corrupt the image, including the padding. In that case you'd have to do decide if you want to keep the ragged edges of the image or not. If you want straight edges even though your edges are not straight then you could count the pixels that match the padding value and if a certain percentage of them match, consider that as a padding row or column. Maybe something like

percentage = 70 / 100; % 70% or whatever.
goodRows = sum(~mask, 2) > percentage * columns;
goodColumns = sum(~mask, 1) > percentage * rows;

[DGM]

That was kind of my point. I would expect something more like this from the first example:

A = imread('testmontage.png');
B = cropborder(A);

Whereas your example method will remove the padding between the images as well, since those vectors also have no non-padding content.

I think a bit of tolerance can help some with noisy edges if the contrast is good (it usually is for figure captures), but I went about it a bit differently due to the images I was dealing with at the time. The default method searches inward, looking for a significant increase in the variance of the image vectors parallel to the image boundary. It's not unbreakable, but that's why cropborder() has other available methods and an adjustable tolerance parameter. It also allows for use on images with obnoxious decorative patterned borders.

[Image Analyst]

Yes, different input images and different desired final appearances will require different code to be run.

For a common case of the user doing File/Save As... to save the figure (including a big white frame around it) as a PNG file, instead of using imwrite() to save the image alone, my original code will work fine.

Answer 7

[yanqi liu]

% I have a grayscale jpg-image, with a very large white space around the actual image, 
% which is dark. Is there an easy command, to remove the colors <250 from the image?
clc; clear all; close all;
im = imread('https://ww2.mathworks.cn/matlabcentral/mlc-downloads/downloads/submissions/34898/versions/8/screenshot.png');
im2 = imcrop(im, [903 133 592 355]);
imt = im2;
im2 = rgb2gray(im2);
im2(im2>250) = median(im2(:));
figure('Color', 'c');
subplot(1,2,1); imshow(imt,[]);
subplot(1,2,2); imshow(im2,[]);