# Text skew detection and correction in image

7 visualizzazioni (ultimi 30 giorni)
Yahya il 19 Nov 2022
Commentato: Image Analyst il 22 Nov 2022
clc;clear all; close all;
figure;
imshow(a)
if size(a,3)==3
a=rgb2gray(a);
end
b=im2bw(a);
ar=imrotate_white(b,30);
figure
imshow(ar);
br=~(ar);
figure
imshow(br)
[r c]=size(b);
X=[];
for i=1:r
for j=1:c
if br(i,j)==1;
X=[X; i j];
end
end
end
size(X);
[coeff, score, latent]=pca(X);
figure;
biplot(coeff(:,1:2),'Scores',score(:,1:2),'VarLabels',{'v_1','v_2'});
r=sqrt((coeff(1,1))^2+(coeff(1,2))^2);
th=real(double(vpa(acos(coeff(1,1)/r)*180/pi)))
syms x
th=real(double(vpa(solve(x==th-90))))
rotatedImage=imrotate(br,-th);
figure
imshow(rotatedImage)
function rotated_image = imrotate_white(image, rot_angle_degree)
RA = imref2d(size(image));
tform = affine2d([cosd(rot_angle_degree) -sind(rot_angle_degree) 0; ...
sind(rot_angle_degree) cosd(rot_angle_degree) 0; ...
0 0 1]);
Rout = images.spatialref.internal.applyGeometricTransformToSpatialRef(RA,tform);
Rout.ImageSize = RA.ImageSize;
xTrans = mean(Rout.XWorldLimits) - mean(RA.XWorldLimits);
yTrans = mean(Rout.YWorldLimits) - mean(RA.YWorldLimits);
Rout.XWorldLimits = RA.XWorldLimits+xTrans;
Rout.YWorldLimits = RA.YWorldLimits+yTrans;
rotated_image = imwarp(image, tform, 'OutputView', Rout, 'interp', 'cubic', 'fillvalues', 255);
end
The output is
Rotate orgin image angle= 30
The answer is angle = 25.1508 !!!!
The angle must be 30 .
How do I solve this problem?
This problem appears to me in the rational equations only, but in the linear equations the angle is shown with high accuracy.
##### 0 CommentiMostra -2 commenti meno recentiNascondi -2 commenti meno recenti

Accedi per commentare.

### Risposte (1)

cr il 19 Nov 2022
Modificato: cr il 19 Nov 2022
In general, pattern maching is used. See estimateGeometricTransform()
But the result may not be exactly the angle it's rotated to.
##### 2 CommentiMostra NessunoNascondi Nessuno
Yahya il 22 Nov 2022
Not useful :( ,thanks for answering
Image Analyst il 22 Nov 2022
@Yahya Then why did you accept the answer?
In general, for an arbitrary, random set of blobs, there is no way to determine what is the "correct" skew and orientation angle.

Accedi per commentare.

### Categorie

Scopri di più su Denoising and Compression in Help Center e File Exchange

### Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by