Using OpenCV in linux/windows/mac os

Open CV

A sample tutorial for images

How does djvu coding compresses scanned images?

Any idea?

Things to learn in the next 1 year

Open CV on linux platform
Awareness of the basic libraries available in C/C++
Mex interface
Running batch process/ parallelizing your algorithm to run on clusters

Some new stuff learnt from MATLAB

1. How to run a peice of code on an array of data and find out the ones for which the function gives a runtime error

try and catch block

try
output(i)=function(intput(i))
catch ME1
% ME1.identifier
fid=fopen("errorfile.txt",'w');
fprintf(fid,'%s\n',ME1.message);
for j=1:length(ME1.stack)
fprintf(fid,'filepath = %s\n',ME1.stack(j).file);
fprintf(fid,'function name = %s\n',ME1.stack(j).name);
fprintf(fid,'line no = %d\n',ME1.stack(j).line);
end
fclose(fid);

end



2. Moving and copying files around in matlab

copyfile('src','dst');
movefile('src','dst');

Need help on vector space projections

I thought it was not required to write my thesis...but need it :(.........Who has any idea of vector space projections used for image retrieval ??? The most quoted text book by Henry Stark is not available in India...
The idea is
a) given |autocorrelation matrix|^2 of an image
b) |spectrum|^2 of the image (say by taking FFT)
can we determine the phase of the image points....so to reconstruct the complete image. (As you know ||^2 takes away all the information about phase.

Any idea of autocorrelation matrix of images ?

For a one-dimensional sequence x[n], the autocorrelation matrix is E(xx'), where E is the expectation operator and x' denotes transpose of x[n]. This matrix has Toeplitz structure. If we extend the case for a two-dimensional sequence x[n,m], what is the structure of resulting matrix ?

Any thoughts ?

Origin of Noise in images

The concept of noise is very intuitive in communication systems. When signals are transmitted through a channel, various sources such as thermal noise, shot noise corrupt the data. The noise is usually modelled as additive white noise with Gaussian statistics.

I want to know what noise sources, if any, originate in the field of image processing. What do we mean when we say (if we do say) that an image is buried in noise? What are the sources?

A related concept is the distortion. Technically both distortion and noise modify the data. However, we can always compensate for the distortion (meaning it is not random..but deterministic) but cannot do so for noise. Blurring of images comes to my mind as source of distortion. Why does blurring occur?

An interesting point regarding 2D z-transforms. In 1D, a polynomial of degree N can always be factored in terms of its zeros, like
f(z) = A(z-z0)(z-z1)(z-z2).....(z-zN).

Surprisingly we can't do this for 2D functions such as f(u,v), where u and v are in general complex. This is supposed to be linked to concept of vector space projections. Any idea what this means ?