growUntil.m

function complete_region=growUntil(g, spIncl, ~, sp_mean, ~, sp_std, sp_rcount, ~)
% complete_region=growUntil(g, spIncl, ~, sp_mean, ~, sp_std, sp_rcount, ~)
%TODO: use a new input: mean of all regions merged in mergeRegionsSimple
%TODO: don't recalculate mean/text each time.  calc new using old...

% V5.1 fixes a bug in calculating region mean vs region index...
% V5 uses proper weighting.  .count refers to pixels
% V4 makes uses of simply-merged regions, so needs a region size input
% V2 use alternative to unique() function and fixes mean bug
% V1 uses dynamic programming
% dilates a superpixel 'image' (graph), g, for the region containing
% labled sps 'SP_incl' (vector), using graph g (of initial water SPs) until
% condition is met
% lim are limits of growing as multiple of the standard deviation of the
% region (sp_std), as given by the global var f.bounds (ex: .9, 1.1).
% note spIncl gives indexes to sp_mean;
% add variance, entropy, or texture image 
% calls SP_dil() and fastSetdiff()
% sp_rcount is vector giving sizes of every SP, in pixels.  Function
% returns a new graph containg SP indexes for the new, dilated regoion.


%%
% init
global f
% dil_sps=[]; %vector
% completeRegion_old=0;
newring.idxs=1.0;
complete_region=spIncl; clear spIncl % units of SP index!
firstTime=true;
% loop
region.std=max(sp_std(complete_region)); % max in case regions not merged
c=0; % counter

%
region.count=sp_rcount(complete_region); % pixel count of region before growing
region.mean=sum(double(sp_mean(complete_region)).*region.count)/...
    sum(region.count);  
%
bounds.a=(region.mean-f.bounds(1)*region.std);
bounds.b=(region.mean+f.bounds(2)*region.std);
%
while newring.idxs~=0
    c=c+1;
        % find indexes of buffered SPs
    ring.idxs=setdiff(SP_dil(g, complete_region), complete_region);
    region.idxs=complete_region; % this is my output dilated variable
        % need to weight mean and std based on number of pixels
%     region.text=sqrt(double((sp_text(region.idxs))).^2.*region.count/...
%         sum(region.count)); % maybe not useful
%     region.var=std(double(sp_mean(ring.idxs))); % maybe not useful
        % count 
        
%     bounds.a=repmat(lims(1)*mean(region.mean), length(ring.idxs),1) ; 
%     bounds.b=repmat(lims(2)*mean(region.mean), length(ring.idxs),1) ; 
    

%     fprintf('Region mean= %f\n', mean_region)
%     fprintf('Length of complete_region= %d\n', length(complete_region))
%     fprintf('Length of newring= %d\n', length(newring))
%     fprintf('bounds= %5.0f\t%5.0f\n', bounds(1).a, bounds(1).b)
%      std_region=std(sp_mean(SP_incl)); % single/double prob...

    newring.idxs=ring.idxs(sp_mean(ring.idxs)>=bounds.a & sp_mean(ring.idxs)<=bounds.b);
    newring.count=length(newring.idxs)*sum(sp_rcount(newring.idxs));
    if length(newring.idxs)>f.maxDilation % regions must have fused, or there was error
        fprintf(' **Dilation > %d regions.  STOP**.', f.maxDilation)
        break
    else
        newring.mean=mean(double(sp_mean(newring.idxs)).*sp_rcount(newring.idxs));
        complete_region=unique([complete_region; newring.idxs]);
    end
    if newring.count>0 % &~firstTime
        fprintf(' %d', length(newring.idxs))
    end
    firstTime=false;
    if c>= f.growMax % probably growing out of control!
        fprintf(' **STOP**')
        newring.idxs=0;
%         pause
        break
    end
%     SP_plot(complete_region, stats, size(L_all)); pause(.01)
end
fprintf(' | Bounds: %3.1f  %3.1f | Mean: %3.1f', mean(bounds.a), mean(bounds.b), region.mean);
% complete_region=unique(complete_region);
% histogram(sp_mean(region.idxs), 'BinMethod', 'integers') %show PMF for superpixels of interest for debugging purposes
% disp('') % for adding a breakpoint