How to histogram equilize an image?

Suppose i have a set of face images I1,I2, ,IM ,size of each image being LxL,in the training set.Now how to histogram equalize all the images and represent them as row vectors of size L^2 (i mean L to the power 2)each...

Answer 1

[Sean de Wolski]

doc histeq
doc reshape

[Manoj Murali]

Error:

??? Error using ==> iptcheckinput

Function HISTEQ expected its first input, I, to be two-dimensional.

Error in ==> histeq at 71

iptcheckinput(a,{'uint8','uint16','double','int16','single'}, ...

Error in ==> histeq1 at 3

J = histeq(I);

the code I wrote is:

clc;

I = imread('C:\Users\Manoj\Documents\MATLAB\sorry.tif');

J = histeq(I);

imshow(I)

figure, imshow(J)

[Jurgen]

Your error is because the matrix you want to equalize is not 2D. Im assuming its an RGB image. Either equalize all color channels separately (dangerous) or convert to a difference colorspace.

If you convert to e.g. L*a*b* or YCbCr you only need to equalize the luma channel, and then convert back again to RGB.

Note you can not represent an LxL RGB image as a vector of length L^2 without resorting to:

1. throwing away data (e.g. color)
2. indexing or some other form of encoding.This is because your RGB image contains LxLx3 elements.

Suppose you convert to grayscale then it is possible by using reshape or just J(:). To avoid columnwise dumping transpose first (permute for N-Dim.) e.g. (J = J').