plus_filt1d.m

适用于matlab环境的源代码

function [data_out, kernel] = plus_filt1D(data_in, x, mu, sigma, sigma_width)
% function to calculate PLUS edge detector with with 1D kernel is
% the special case of normal 2D PLUS filter (see plus_filt2D.m)
% Depending on syntax, it uses either 1D gaussian kernel or a linear derivative
% kernel in form of [-1 N*0 1], where N is the odd number of zeros, or any other
% kernel supplied through the variable X. For linear case only one (DATA_IN) or two 
% input variables (DATA_IN and X)are used and in this case X must be a vector.
% First and second derivative kernels are the same size
%
%INPUTS:
%   DATA_IN - input vector or array
%   X       - vector in form of [-1 N*0 1] containing derivative kernel in linear 
%           case or number of points for Gaussian kernel. In case when only two input
%           variables exist (DATA_IN and X) length of the X determines which kernel 
%           will be used - linear (X==vector) or Gaussian (X==scalar) 
%   MU      -  mean of the gaussian
%   SIGMA   - standard deviation of a gaussian function
%           sigma_width- defines where to cut the Gaussian kernel tails (or width of 
%           the kernel in sigma). Bigger number- more of Gaussian included in the kernel.
%           Defaul value= 3 or 98% of Gaussian 
%
%OUTPUTS:
%   DATA_OUT- filtered data output
%   KERNEL - first derivative kernel used for filtering
%
%WARNING:   Derivatives for some pixels may be zero. This produce an error "divideByZero" 
%           and such pixels gets value =NaN. It is recomended to check for NaNs after 
%           filtering and replacing it with value of convenience -> zero, global minimum, etc.
%           To find all NaNs, Inf and -Inf and set them to zero you may use the following line:
%           data_out(~isfinite(data_out)) = 0;
%
%EXAMPLES:
%X = imread('tire.tif'); imagesc(X); colormap gray;
%%This image can be found at \MATLAB704\toolbox\images\imdemos
%
%%Try one of the following:
%[dd, kernel] = plus_filt1D(X);  dd = dd ./(max(dd(:)));
%figure; imagesc(dd> 0.01); colormap gray;
%[dd, kernel] = plus_filt1D(X, [-1 0 0 0 0 0 0 0 0 0 1]); dd = dd ./(max(dd(:)));
%figure; imagesc(dd> 0.01); colormap gray; colorbar
%
%%OR
%
%[dd, kernel1] = plus_filt1D(X, 11, 0, 1, 3); %1
%dd = dd ./(max(dd(:))); figure; imagesc(dd> 0.1); colormap gray; 
%[dd, kernel2] = plus_filt1D(X, 15, 0, 1, 3); %2
%dd = dd ./(max(dd(:))); figure; imagesc(dd> 0.1); colormap gray; 
%[dd, kernel3] = plus_filt1D(X, 21, 0, 1, 3); %2
%dd = dd ./(max(dd(:))); figure; imagesc(dd> 0.1); colormap gray; 
%figure; plot(kernel1, 'r.-'); hold on; 
%plot(kernel2, 'b.-'); plot(kernel3, 'k.-'); grid; axis tight
%
%REF: 
%1   Marr D., Hildreth E.C. in  揟heory of edge Detection