Image correspondences using cross-correlation
16 Apr 2010
09 Dec 2011)
Find matching features in pairs of images using normalised cross-correlation: class file and demo.
|corrpeak(i, m, n, tol)
function [r,c,v] = corrpeak(i, m, n, tol)
%CORRPEAK Find peak of correlation.
% [R, C, V] = CORRPEAK(I, M, N, TOL) Return the R,C coordinates in I of
% the peak of its correlation with M, together with the peak correlation
% value V. TOL is the relative accuracy with which M is represented in
% the correlation. The centre point of M is taken as its (0,0) coordinate
% point, so if M has even size R and/or C may not be a whole number.
% N is a normalisation matrix - the correlation result is divided by it
% before the peak is found. It needs to be the same size as a 'valid'
% convolution of i with m.
% TOL may be omitted and defaults to 0. Both TOL and N may be omitted:
% TOL then defaults to 0 and no normalisation is applied. If N is  then
% likewise no normalisation is used.
% Copyright David Young 2010
if nargin < 4; tol = 0; end
if nargin < 3; n = ; end
m = rot90(m, 2); % convert to convolution
cc = convolve2(i, m, 'valid', tol);
if ~isempty(n); cc = cc ./ n; end
[v,r,c] = max2(cc);
% Allow for offset caused by taking valid region
offsets = (size(m)-1)/2;
r = r + offsets(1);
c = c + offsets(2);