Problems with segments in an image

Question

Harold 2013-4-21

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/72971-problems-with-segments-in-an-image

I'm working on a project that involves character recognition. If I load an image that has for example the letters A, B, C, D, my program finds the regions and does some operations on it to get rid of vertical and horizontal white lines and also re-sizes the image to a pre-defined size. This works fine. However, if I load an image up that has one letter for example A, the bwlabel segments the image into 4 regions because that's the minimum number of regions it can have. I need to be able to code some logic that will help with this.

Here is the procedure that I have developed thus far (it has problems). num is the number of regions in the image from bwlabel. Ibw is the logical matrix for the image. Iprops is a structured array of properties for each region. Each row is a region.

grid_size = [10, 10];
[r c] = size(Iprops);
if num == 4 && num == r             %single character recognition
    segments = imresize(Ibw, grid_size, 'bicubic');
    segments = segments(:)';
end

The problem with this code lies in the use of num == r. An image of one letter will return the same number of properties as an image of 4 letters. I need to modify my logic to account for this.

Edit: Now that I think of it, if I can determine if any regions overlap I could include that in my logic statement. The questions arise though, how can I check if any regions overlap?

3 个评论
显示 1更早的评论隐藏 1更早的评论

Harold 2013-4-22

编辑：Harold 2013-4-22

I tried the region overlap idea but realized that will not work out because the matrices of each region are a different size.

http://i33.tinypic.com/1zdrneb.jpg

http://i36.tinypic.com/2u9igpt.jpg

Harold 2013-4-22

I've re-uploaded both to a different hosting site.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Image Analyst 2013-4-22

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/72971-problems-with-segments-in-an-image#answer_83025

在 MATLAB Online 中打开

There is no reason why the A would be split up into 4 regions. I'd have to see your code. Do you do something like

blackLetters = grayImage < 128;
[labeledImage, numberOfBlobs] = bwlabel(blackLetters)

???

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Harold 2013-4-22

编辑：Harold 2013-4-22

在 MATLAB Online 中打开

Not exactly. Here is what I have. Image file is http://i37.tinypic.com/29err4w.jpg

clc
clear all
filename = 'A.bmp';
I = imread(filename);
imshow(I)
%convert image from rgb to gray scale
Igray = rgb2gray(I);
%convert grayscale image to binary image
Ibw = im2bw(Igray,graythresh(Igray));
%Find edges in grayscale image
Iedge = edge(uint8(Ibw));
%Image Dilation (thickens the edges)
se = strel('square',2);
Iedge = imdilate(Iedge, se);
%Image Filling - gets rid of any background pixels inside the segments
Ifill= imfill(Iedge,'holes');
[Ilabel num] = bwlabel(Ifill);
Iprops = regionprops(Ilabel);
Ibox = [Iprops.BoundingBox];
Ibox = reshape(Ibox,[4 num]);
%Draw bounding box around segmentations
for cnt = 1:num
    rectangle('position',Ibox(:,cnt),'edgecolor','r');
    point =  Iprops(cnt,1).Centroid;
    text(point(1),point(2),num2str(cnt),'HorizontalAlignment','center','BackgroundColor',[.7 .9 .7]);
end