Detecting text from image with perspective distorted using matlab
1 answer
I think matlab is good for text detection (ocr). There are several libraries for ocr. The most important ones are:
- Abbyy
- Tesseract
Tesseract is considered to be one of the most accurate open source open source OCR engines currently available.
I have been working on a perspective distortion ocr and I am using the matlab-tesseract library.
I have an issue with perspective distorting perspective with image rotation by finding the border of the image and its lines using the houghlines method .
I will share my code snippet and hope it gives you an idea of ββthis problem.
function [ output_args ] = rotate_image( )
clc;
close all;
clear;
I = imread("any_images")
I2 = imcrop(I,[85 150 590 480]);
I3 = imcrop(I2,[-85 0 440 480]);
imwrite (I3,'original.tiff', 'Resolution', 300);
thresh = graythresh(I3);
Bimage = im2bw(I3, thresh);
Bimage = edge(Bimage,'sobel');
[H, T, R] = hough(Bimage);%,'RhoResolution',0.1,'Theta',[-90,0]);
P = houghpeaks(H,10);%,'threshold',ceil(0.3*max(H(:))));
% Find lines and plot them
lines = houghlines(Bimage,T,R,P);%,'FillGap',100,'MinLength',5);
figure, imshow(I3), hold on
max_len = 0;
for k = 1:length(lines)
xy = [lines(k).point1; lines(k).point2];
len = norm(lines(k).point1 - lines(k).point2);
if ( len > max_len)
max_len = len;
xy_long = xy;
end
end
% highlight the longest line segment
plot(xy_long(:,1),xy_long(:,2),'LineWidth',2,'Color','blue');
a = -xy_long(1,1) + xy_long(2,1);
b = -xy_long(1,2) + xy_long(2,2);
angle=180-atand(b/a);
if angle > 180
angle = angle-180;
end
angle
B = imresize(I3, 2);
if angle > 90
B = imrotate(B, 90 - angle ,'bilinear','crop');
else
B = imrotate(B, -angle,'bilinear','crop');
end
imwrite (B,'rotated_image.tiff', 'Resolution', 300);
end
+2
source to share