Filling a Bounding Box with zeros

I am trying to Replace an boxed region with all 0's in a binary image

BW2 is the image & the object lies at

[642.5000 1.5000 55.0000 82.0000] (from regionprops-BoundingBox)

BW2(642.5:697.5,1.5:82.5)=zeros(55,82);-gives"Subscripted assignment dimension mismatch."

BW2(642.5:697.5,1.5:82.5)=0;-Doesn't replace the regions with zeros

BW2(642.5:697.5,1.5:82.5)=[0];-Doesn't replace the regions with zeros

How to do this Change all the pixels to zero with in that box??

Answer 1

[Image Analyst]

You can't use fractional indexes. Keep in mind that bounding box goes outside the pixels, and the pixel has it's center at integer indexes, so to be outside, you need to be a half pixel away from the center. So you need to do ceil() on the left and top, and do floor() on the right and bottom edges.

The main problem is that you made the very common beginner mistake of thinking (x,y) = (row, column). It does not. The bounding box from regionprops() returns [x, y, width, height], NOT [row, column, height, width]. You took the first argument x (the horizontal/column location) and used it as the vertical/row/first index instead of the horizontal/column/second index. You can't do that. You can do it correctly this way:

bbox = props.BoundingBox;
x1 = ceil(bbox(1));
x2 = round(x1 + bbox(3));
y1 = ceil(bbox(2));
y2 = round(y1 + bbox(4));
BW2(y1:y2, x1:x2) = 0; % NOTE : y comes first, NOT x!!!

Note that what you did was essentially

BW2(x1:x2, y1:y2) = 0; % This is wrong.

[akpojotor princewill]

I followed this pattern. but when i display BW2 it doesn't show that the bounded regions have been replaced with 0's. How can I fix that

[Image Analyst]

Start your own new question and attach your code and image.

[Shigeru Kato]

As well, I confuse.

Even though size of image or anchor box is defiined as [row,col] i.e. [height, widht],

On the other hand, a rectangle form is defined as [x,y, width, height].

It is annoying me too much.

[Image Analyst]

@Shigeru Kato, yes I understand and share your pain. The (x,y)/(row/column) issue is something you always have to keep in mind. It's not just images, it's also a watchout for matrices in general.

Also the [x,y,width,height] used in rectangle() and imcrop() and other functions can sometimes be less convenient than if we had a [x1,y1,x2,y2] or [x1,x2,y1,y2] option.

[Shigeru Kato]

@Image Analyst

Thank you very much for your response.

I made a script for confirmation of my understanding.

clear;
clc;
close all;
    
A = imread('peppers.png');
[row, col, dim] = size(A)

imshow(A);
hold on;

x = 50;
y = 300;

plot(x, y, '*b');
text(x,y, '  [x,y]', 'Color', 'b');

plot([x,x],[0,y], '--c');
plot([0,x],[y,y], '--g');

hold off;

[Marco A. Acevedo Z.]

I would like to point out that using the technique outline above is the best option. However, if using imcrop() with the bounding box, you will have 1 more pixel (in columns) for some reason. Maybe, the rounding of .5 values is different in the inputs of imcrop() since I had converted strings to double to define the bounding box. So, I recommend the solution of Image Analyst. Bye.

[Image Analyst]

@Marco A. Acevedo Z., imcrop() is meant to have rectangles, like the boundingBVox returned from regionprops(), that are a half pixel outside the region to be cropped. It's not meant to work on integer indexes. I know it's confusing but that's the way it is. If you want to crop using integer indexes, simply using indexing, not imcrop().

Answer 2

[Anand]

Why does this not work?

BW2(642:642+56,1:1+83) = 0;

Give us something more than "Doesn't replace the region with zeros".

[Vijay]

it gives a larger image than the input image 1536X2048 becomes 2048X2048 and that also filled filled improperly column wise improperly.. so i have started using ismember function for the condition and to remove the objects not satisfying the condition for text extraction using the bounding box criteria..

[Julio Cesar Garcia Navarro]

Hello,

I tried the:

BW2(642:642+56,1:1+83) = 0;

And it indeed changes the size of the image; however, if you transpose the values, you will get the correct Bounding Box filled without modifying the image size. Thus:

BW2(1:1+83,642:642+56) = 0;

Will be correct.

Answer 3

[VINSHI K K]

Hello, I need to fill bounding box region with ones..could you please suggest the code.

[kanika bhalla]

Hi @VINSHI K K

I am not sure whether you need to know this or not. But I was just looking at this post. So, read your comment.

Here is a solution which I tried now.

Colorimage=imread('1.jpg')
Binaryimage=imread(binary.jpg) % C is a binary image 
info = regionprops(binary,'Boundingbox') ;
hold on
for k = 1 : length(info)
     BB = info(k).BoundingBox;
     rectangle('Position', [BB(1),BB(2),BB(3),BB(4)],'EdgeColor','r','LineWidth',2) ;   
    x= BB(1)
    y=BB(2)
   width= BB(3)
   height=  BB(4)
end
hold off; 
figure; imshow(Colorimage); % if you want same bounding box in colored image
x= BB(1)
y=BB(2)
width= BB(3)
height=  BB(4)
hold on;
rectangle('Position',[x,y,width,height],...
  'EdgeColor', 'r',...
  'LineWidth', 3,...
  'LineStyle','-')
hold off;
% Below is the method to make your bounding box object to 255.
 x1 = ceil(x);
x2 = round(x1 + width);
y1 = ceil(y);
y2 = round(y1 + height);

Image(y1:y2,x1:x2,:) = 255 % Note if your image is RGB
Image(y1:y2,x1:x2) = 255 % if your image is binary. I mentioned this because to solve this error costs me 7 hours of debugging. 

[Image Analyst]

@kanika bhalla,I think you meant this for setting the pixels in your images:

Colorimage(y1:y2, x1:x2, :) = 255; % Note if your image is 3-D and RGB
Binaryimage(y1:y2, x1:x2) = true; % if your image is 2-D binary (logical). I mentioned this because to solve this error costs me 7 hours of debugging. 

Also, it's probably not a good idea to name your vairable Image since there is a built in function called image(). True, MATLAB is case sensitive so it wouldn't overwrite the image functions, but nonetheless it's probably best not to have variables with names like functions. Also your code finds (potentially) many bounding boxes but you only assign the last one to white. If you want the others, then that code should be inside your for loop.

[kanika bhalla]

Thank you so much Sir @Image Analyst. Everyday I am learning a lot from you and Matlab community. I am really grateful.