function permutedFeatureVectors = makeAppendedRelativeFeatures(rownumber, image, scales)
% Version 0.1, Feb 2006.
% This code is copyrighted by Ashutosh Saxena, Stanford University
% It is available for non-commercial use only. For commericial use, please
% contact the authors.
% Any publication or report resulting from Use of this code, should cite:
% Learning Depth from Single Monocular Images, Ashutosh Saxena, Sung H. Chung, Andrew Y. Ng, NIPS 2005.
% 3-D Depth Reconstruction from a Single Still Image, Ashutosh Saxena, Sung H. Chung, Andrew Y. Ng, To appear in IJCV 2007.
% For more info and dataset, visit: http://ai.stanford.edu/~asaxena/learningdepth/
% For a better version of the algorithm, also visit: 
% http://ai.stanford.edu/~asaxena/reconstruction3d/



global nGridRow nGridCol featureDirectory
global nStatistics

configurationFile;

numStatistics = nStatistics;
% Per scale
numfeatures = 170;
% Scales must follow the order of scales in the mat file
numscales = length(scales);

load([featureDirectory 'allTrainingRelativeFeatureData_row' ...
	num2str(rownumber) '_' num2str(image) '.mat']);
currowFeatureVector = rowRelativeFeatureVector;
numimages = size(currowFeatureVector, 4);
numcolumns = size(currowFeatureVector, 2);

numNeighbors = 3;
% middle, bot, right order
permutedFeatureVectors = zeros(1, numcolumns, ...
	numNeighbors*numscales*numfeatures, numimages);

% Let's load all the necessary feature vectors
for z=1:numscales
	curscale = scales(z);
	k = 0;
	if curscale == 1
		k = 1;
	elseif curscale == 2
		k = 2;
	elseif curscale == 4
		k = 3;
	else
		display('Should not get here !');
	end

%	offset = 1;
%	if curscale == 2
%		offset = (curscale - 1)*2 + 1;
%	end
	offset = 3;
	botrow = rownumber + offset;
	if botrow > nGridRow
		botrow = nGridRow;
	end

	% The feature at the center
	permutedFeatureVectors(1, :, ...
		(numNeighbors*numfeatures*(z-1)+1):(numNeighbors*numfeatures*(z-1)+numfeatures), :) = ...
		currowFeatureVector(1, :, ((k-1)*numfeatures+1):(k*numfeatures), :);

	% The feature at the bottom
	load([featureDirectory 'allTrainingRelativeFeatureData_row' num2str(botrow) '_' ...
		num2str(image) '.mat']);
	permutedFeatureVectors(1, :, ...
		(numNeighbors*numfeatures*(z-1)+numfeatures+1):(numNeighbors*numfeatures*(z-1)+2*numfeatures), :) = ...
		rowRelativeFeatureVector(1, :, ((k-1)*numfeatures+1):(k*numfeatures), :);

	% The feature to the right
	perm = zeros(1, numcolumns);
	for h=1:numcolumns
		if (h+offset) > numcolumns
			perm(1,h) = numcolumns;
		else
			perm(1,h) = h + offset;
		end
	end
	permutedFeatureVectors(1, :, ...
		(numNeighbors*numfeatures*(z-1)+2*numfeatures+1):(numNeighbors*numfeatures*(z-1)+3*numfeatures), :) = ...
		currowFeatureVector(1, perm, ((k-1)*numfeatures+1):(k*numfeatures), :);
end