function [ outputImg ] = isolatePlate( inputImg )
% ISOLATEPLATE Draw a box around a license plate in the input image
%  More info.



% Params
%SIZE_THRESHOLD = 60;
TRUE = 1;
FALSE = 0;


% Show the original
inputImg = im2double( inputImg );
%figure, imshow( inputImg ), title( 'Original Image' );



%%%%% Image Prep

% Convert To Grayscale Using the NTSC standard

red = immultiply( inputImg(:,:,1), 299 );
green = immultiply(  inputImg(:,:,2), 587 );
blue = immultiply( inputImg(:,:,3), 114 );

newImg = red + green + blue;
newImg = imdivide( newImg, 1000 );



% Smooth the image
smoother = ones( 6, 6 );
newImg = imfilter( newImg, smoother );
newImg = imdivide( newImg, 36 );

%figure, imshow( newImg ), title( 'Grayscale Image' );

flipImg = rot90( newImg );








% create a new light gray image so we can see the boxes.
lightGray = imdivide( newImg, 4 );
lightGray = imadd( lightGray, 0.749 );

% % Find the center of the red data, to find the column that the license
% % plate should cross
% 
% scaledRed = mat2gray( red );
% scaledBlue = mat2gray( blue );
% scaledGreen = mat2gray( green );
% 
% % threadhold the red values...
% redMap = scaledRed > 0.30 & scaledGreen < 0.40 & scaledBlue < 0.40;
% 
% se = strel( 'disk', 5 );
% redMap = imclose( redMap, se );
% 
% [row,column] = find( redMap );
% redColumn = sum( column ) / size( column, 1 );
% redRow = sum( row ) / size( row, 1 );
% 
% figure, imshow( imcomplement( redMap ) ), title( 'Red Map' );
% rectangle( 'Position', [ redColumn 1 1 size( red, 1 ) ] );
% rectangle( 'Position', [ 0 redRow size( red, 2 ) 1 ] );

%%%%% Vertical Edge Detection

% Create our mask - a Sobel for vertical edge detection
mask = fspecial( 'sobel' );
mask = rot90( mask, 1 );

% Do the actual detection
vEdges = imfilter( newImg, mask );
hEdges = rot90( imfilter( flipImg, mask ), -1 );

vEdgesScaled = immultiply( vEdges, 0.7 );
hEdgesScaled = immultiply( hEdges, 0.3 );
edges = imadd( vEdgesScaled, hEdgesScaled );

% Show what we have so far
%figure, imshow( vEdges ), title( 'Vertical Edges' );
%figure, imshow( hEdges ), title( 'Horizontal Edges' );
%figure, imshow( edges ), title( 'Edges' );

%%%%% Edge Density Map Generation

% Create our mask of ones to add up the edge values
%densityMask = ones( 3, 15 );
densityMask = ones( 3, 15 );

% Sum it up
edgeMap = imfilter( edges, densityMask );
edgeMap = imdivide( edgeMap, 45 );

% Show what we have so far
%figure, imshow( edgeMap ), title( 'Edge Map' );



%%%%% Binarization / Clean Up / Dilation

% Binarize!!!!  Use Otsu's Method.  It's fast.
level = graythresh( edgeMap );
binary = im2bw( edgeMap, level );

% Show what we have so far
%figure, imshow( binary ), title( 'Binarized Edge Map' );

% Clean Up by getting rid of small horizontal lines

se = strel( 'disk', 2 );
cleanImage = imopen( binary, se );

% Show what we have so far
%cleanImage = binary; % skipping the clean up step...don't seem to need it
%figure, imshow( cleanImage ), title( 'Cleaned Image' );

% Dilate the Image
%se = ones( 1, 9 );
%se = ones( 100, 100 );
se = strel( 'disk', 13 );
dilated = imdilate( cleanImage, se );

% Show what we have so far
%figure, imshow( dilated ), title( 'Dilated Image' );



%%%%% License Plate Location

% find all the 8-connected regions in the dilated binary image
[labeled, num] = bwlabel( dilated );
%figure, imshow( lightGray ), title( 'Original Image - All Connected Regions' );
figure, imshow( inputImg ), title( 'Original Image - All Connected Regions' );

% Feature Extraction
% here we will loop through the dilated image...
info = [];
imgWidth = size( newImg, 2 );
imgHeight = size( newImg, 1 );

for k = 1:1:num
    % get the indices of one connected region (current k value for this
    % iteration)
    [row,column] = find( labeled == k );
    
    % get the top left row&column
    topLeftRow = min( row );
    topLeftColumn = min( column );
    
    % BR row&column
    bottomRightRow = max( row );
    bottomRightColumn = max( column );
    
    % get the size of...yeah...
    numPixels = size( row, 1 );
    
    % numbers!
    width = bottomRightColumn - topLeftColumn + 1;
    height = bottomRightRow - topLeftRow + 1;
    
    % Check whether the box is within another box
    
    % structure of each row: TL row, TL col, BR row, BR col, numpixels,
    % width, height.  Note that it's 7 columns...
    newRow = [ topLeftRow topLeftColumn bottomRightRow bottomRightColumn numPixels width height ];
    

    % add it to our dater structure to store it!
    ratio = width / height;
    area = width * height;
    
    % check aspect ratio
    if ( ratio > 1.3 & ratio < 2.35 )
        % check min. size through area
        if ( area > 16000 )
            % check max size through percent of image
            if ( width < 0.7 * imgWidth & height < 0.7 * imgHeight )
                info = cat( 1, info, newRow );
                rectangle( 'Position', [ min( column ) min( row ) width + 1 height + 1 ] );
            end
        end
    end
end


% Combination of Candidate Regions

%figure, imshow( imcomplement( inputImg ) ), title( 'Original Image - Regions To Connect' );
%figure, imshow( lightGray ), title( 'Original Image - Regions To Connect' );
densityV = [];




















% looop through all of the regions have tracked
% for oneIndex = 1:1:size( info, 1 )
%     % compare each row to the one after it, so loop over all rows after
%     % this one...like that handshaking problem thingy where everyone has to
%     % shake everyone else's hand.
%     for twoIndex = (oneIndex + 1):1:size( info, 1 )
%         % get the two items to compare....
%         oneInfo = info( oneIndex, : );
%         twoInfo = info( twoIndex, : );
%         
%         % Check whether they're overlapping
%         overlap = TRUE;
%         
%         % we use these cases to determine if the boxes overlap.  If all of
%         % the points are outside, and if it's not than there is an overlap.
%         case1 = oneInfo( 1 ) + oneInfo( 7 ) < twoInfo( 1 );
%         case2 = oneInfo( 1 ) > twoInfo( 1 ) + twoInfo( 7 );
%         case3 = oneInfo( 2 ) + oneInfo( 6 ) < twoInfo( 2 );
%         case4 = oneInfo( 2 ) > twoInfo( 2 ) + twoInfo( 6 );
%         
%         % find out what we have here...
%         if case1
%             overlap = FALSE;
%         elseif case2
%             overlap = FALSE;
%         elseif case3
%             overlap = FALSE;
%         elseif case4
%             overlap = FALSE;
%         else   
%             overlap = TRUE;
%         end
%         
%         % If they overlap, check their connected density
%         if overlap
%             % overlapping region boundaries - vertical...
%             topRow = min( oneInfo( 1 ), twoInfo( 1 ) );
%             insideTop = max( oneInfo( 1 ), twoInfo( 1 ) );
%             insideBottom = min( oneInfo( 3 ), twoInfo( 3 ) );
%             bottomRow = max( oneInfo( 3 ), twoInfo( 3 ) );
%             
%             % vertical overlap...
%             overlap = ( bottomRow - topRow + 1 ) - ( bottomRow - insideBottom ) - ( insideTop - topRow );
%                        
%             % find the overlapping region's boundaries...horizontal...
%             leftColumn = min( oneInfo( 2 ), twoInfo( 2 ) );
%             insideLeft = max( oneInfo( 2 ), twoInfo( 2 ) );
%             insideRight = min( oneInfo( 4 ), twoInfo( 4 ) );
%             rightColumn = max( oneInfo( 4 ), twoInfo( 4 ) );
%             
%             % horizontal overlap...
%             dis = ( rightColumn - leftColumn + 1 ) - ( insideLeft - leftColumn ) - ( rightColumn - insideRight );
%             
%             % find the 
%             minH = min( oneInfo( 7 ), twoInfo( 7 ) );
%             sqr = sqrt( oneInfo( 6 ) * twoInfo( 6 ) );
%             
%             density = ( overlap / minH ) - ( dis / sqr );
%             
%             % thresholding for density...
%             if ( density > 0.7 )
%                 % note: we have an error here when height is 0.  Height
%                 % shouldn't be zero though :-(
%                 r1 = rectangle( 'Position', [ oneInfo( 2 ) oneInfo( 1 ) oneInfo( 6 ) oneInfo( 7 ) ] );
%                 r2 = rectangle( 'Position', [ twoInfo( 2 ) twoInfo( 1 ) twoInfo( 6 ) twoInfo( 7 ) ] );
%                 
%                 r1
%                 r2
%                 
%                 r1.EraseMode = 'xor';
%                 r2.EraseMode = 'xor';
%                 
%             end
%             
%             % keep track of the densities...
%             densityV = cat( 1, densityV, density );
%         end
%     end
% end

% output these values for testing...
% densityV
% max( densityV )
% min( densityV )

% Show what we have so far...
%figure, imshow( labeled ), title( 'Regions of Binary Image' );

% create final image
% final = zeros( size( inv, 1 ), size( inv, 2 ) );
% for k = 1:1:size( iregions )
%    final = final | ( labeled == iregions( k ) );
% end

% todo: Hough transform!

% now return the processed image...
outputImg = inputImg;