bin2mat running slow need a faster solution - matlab

Question

I am using bin2mat function from matlab file exchange, for some reason it runs very slow. Is it possible to make it run faster or is there an alternative? I am trying: zC = bin2mat(s.X,s.Y,s.Z,xC,yC); I am not sure where it gets bogged down. I need to do this on point cloud data to calculate volume and such.
Here is the code:
function ZG = bin2mat(x,y,z,XI,YI,varargin)
% BIN2MAT - create a matrix from scattered data without interpolation
%
% ZG = BIN2MAT(X,Y,Z,XI,YI) - creates a grid from the data
% in the (usually) nonuniformily-spaced vectors (x,y,z)
% using grid-cell averaging (no interpolation). The grid
% dimensions are specified by the uniformily spaced vectors
% XI and YI (as produced by meshgrid).
%
% ZG = BIN2MAT(...,#FUN) - evaluates the function FUN for each
% cell in the specified grid (rather than using the default
% function, mean). If the function FUN returns non-scalar output,
% the output ZG will be a cell array.
%
% ZG = BIN2MAT(...,#FUN,ARG1,ARG2,...) provides aditional
% arguments which are passed to the function FUN.
%
% EXAMPLE
%
% %generate some scattered data
% [x,y,z]=peaks(150);
% ind=(rand(size(x))>0.9);
% xs=x(ind); ys=y(ind); zs=z(ind);
%
% %create a grid, use lower resolution if
% %no gaps are desired
% xi=min(xs):0.25:max(xs);
% yi=min(ys):0.25:max(ys);
% [XI,YI]=meshgrid(xi,yi);
%
% %calculate the mean and standard deviation
% %for each grid-cell using bin2mat
% Zm=bin2mat(xs,ys,zs,XI,YI); %mean
% Zs=bin2mat(xs,ys,zs,XI,YI,#std); %std
%
% %plot the results
% figure
% subplot(1,3,1);
% scatter(xs,ys,10,zs,'filled')
% axis image
% title('Scatter Data')
%
% subplot(1,3,2);
% pcolor(XI,YI,Zm)
% shading flat
% axis image
% title('Grid-cell Average')
%
% subplot(1,3,3);
% pcolor(XI,YI,Zs)
% shading flat
% axis image
% title('Grid-cell Std. Dev.')
%
% SEE also RESHAPE ACCUMARRAY FEVAL
% A. Stevens 3/10/2009
% astevens#usgs.gov
%check inputs
error(nargchk(5,inf,nargin,'struct'));
%make sure the vectors are column vectors
x = x(:);
y = y(:);
z = z(:);
if all(any(diff(cellfun(#length,{x,y,z}))));
error('Inputs x, y, and z must be the same size');
end
%process optional input
fun=#mean;
test=1;
if ~isempty(varargin)
fun=varargin{1};
if ~isa(fun,'function_handle');
fun=str2func(fun);
end
%test the function for non-scalar output
test = feval(fun,rand(5,1),varargin{2:end});
end
%grid nodes
xi=XI(1,:);
yi=YI(:,1);
[m,n]=size(XI);
%limit values to those within the specified grid
xmin=min(xi);
xmax=max(xi);
ymin=min(yi);
ymax=max(yi);
gind =(x>=xmin & x<=xmax & ...
y>=ymin & y<=ymax);
%find the indices for each x and y in the grid
[junk,xind] = histc(x(gind),xi);
[junk,yind] = histc(y(gind),yi);
%break the data into a cell for each grid node
blc_ind=accumarray([yind xind],z(gind),[m n],#(x){x},{NaN});
%evaluate the data in each grid using FUN
if numel(test)>1
ZG=cellfun(#(x)(feval(fun,x,varargin{2:end})),blc_ind,'uni',0);
else
ZG=cellfun(#(x)(feval(fun,x,varargin{2:end})),blc_ind);
end
It is slower on these two steps for one run it took:
ZG=cellfun(#(x)(feval(fun,x,varargin{2:end})),blc_ind); took 33 secs
blc_ind=accumarray([yind xind],z(gind),[m n],#(x){x},{NaN}); took 10 secs

Answer 1

You can change blc_ind = ... to
ZG=accumarray([yind xind],z(gind),[m n],#mean,NaN);
and delete other codes form here so the is no need to if numel(test)>1....