Back to Gallery
  • Remix
  • Share
  • New Entry
on 19 Nov 2023
  • 41
  • 142
  • 0
  • 8
  • 1998
drawframe(1);
Animated version of my 2022 minihack submission Rain On The Window Pane but this one has the following more-realistic raindrop properties:
  • Only the larger window raindrops fall down the window
  • When a window raindrop intersects with another drop, the larger drop absorbs the smaller drop
  • As the falling raindrops increase in size, they are stretched vertically and become narrower horizontally
  • As the falling raindrops increase in size, they fall faster
Did I capture that the raindrops are tickling down on glass or does this just look like rain falling (not on glass)? I can't tell but I sure did spend some time trying 😃 🌧️
function drawframe(f)
persistent T dropSzFcn dropFcn
if f==1 || isempty(T)
rng(359,'twister') % set the random number generator for reproducibility
nc=50; % number of colors in colormap
cmap=gray(nc); % used to set the background color
axes(Position=[0 0 1 1]) % example axes to full size of figure
colormap(cmap(1:ceil(0.4*nc),:)) % cut out the upper (lighter) part of the colormap
hold on
% background image
bx=1:50; % set size of the square image
by=exp(-(((bx-6).^2)/(2*8.^2))); % gaussian
I=imagesc([0,1],[0,1],by'+0*bx);
axis tight equal off
% add distant lights
ncl=99; % number of colors for the lights
LC=hot(ncl); % light colors
nl=30; % number of lights
r=@()rand(nl,1); % generate nlx1 random numbers 0:1
b=bubblechart(r(),r(),r(),LC(randi(ncl,nl,1),:),MarkerEdgeColor='n'); % street lights
figSz=get(gcf,'Pos');
bubblesize([3,0.1*max(figSz(3:4))]) % Scale light size to figure size
% blur the image
% the background and lights are flattened into a single image and
% blurred.
f=getframe;
b.delete
I.CData=imfilter(flipud(f.cdata),ones(10)/100);
% Add rain drops
nd=99; % number of drops
T=table(); % use a table to store some variables
T.obj=gobjects(nd,1); % droplet surf objects
T.dropSz=nan(nd,1); % scale factor for drop size
dropSzFcn=@()max(min(randg(1),6),.8)/150; % Drop size, truncated gamma distribution
[x,y,z]=sphere(20); % use a larger number for smoother raindrop surfaces, but slower.
dropFcn=@(sz)surf(sz*x+rand,2*sz*y+rand*1.1,sz*max(z,0),... % function to create raindrops
FaceCo='w',FaceAl=.2,EdgeCo='n',...
SpecularSt=1,SpecularExp=2, ...
DiffuseSt=1,AmbientSt=.1);
for i=1:nd % Create the rain drops
T.dropSz(i)=dropSzFcn();
T.obj(i)=dropFcn(T.dropSz(i));
end
light(Po=[0.5 -1 0.1]); % rain drops should be dark on top and light on bottom
xlim([0,1])
ylim([0,1])
set(gcf,Color='k')
end
% Add new drops
n=5; % number of rain drops to add
T2=table(); % create a temporary table to store variables
T2.obj=gobjects(n,1); % droplet surf objects
T2.dropSz=nan(n,1); % scale factor for drop size
for k=1:n % add more raindrops
T2.dropSz(k)=dropSzFcn();
T2.obj(k)=dropFcn(T2.dropSz(k));
end
T=[T;T2];
% Determine which rain drops are falling by drop size (larger ones fall)
% figure(); histogram(T.dropSz) % for decision making
T.isFalling=T.dropSz > 0.01; % Reduce threshold to increase the number of falling rain drops
% The amount of downward displacement is determined by drop size
for j=find(T.isFalling')
T.obj(j).YData=T.obj(j).YData-T.dropSz(j); % shift downward
end
% Determine if any drops overlap
% Reduce the computational expense by assuming drops are rectangular and
% useing MATLAB's rectint, though it contains 1 extra step that isn't needed
% (computing area of overlap) but it's still fast and clean.
[mmy(:,1),mmy(:,2)]=arrayfun(@(h)bounds(h.YData,'all'),T.obj); % [min,max] for ydata
[mmx(:,1),mmx(:,2)]=arrayfun(@(h)bounds(h.XData,'all'),T.obj); % [min,max] for xdata
% Covert the drop's x and y data to rectangular vectors [x,y,width,height]
T.xywh=[mmx(:,1),mmy(:,1),diff(mmx,1,2),diff(mmy,1,2)];
% If a water drop is off the figure, remove it
T.isoff=mmy(:,2) < 0;
T.obj(T.isoff).delete;
T(T.isoff,:)=[];
% Compare all pairs of drops without duplicate comparisons
objPairs=nchoosek(1:height(T),2);
overlap=false(height(objPairs),1);
for q=1:height(objPairs)
% Because we're treating the raindrops as rectangles, there will be
% falsely labeled overlaps in the corner of the rectangles. To reduce
% the number of false positives, we'll require the overlap to be at least
% 21.5% of the smallest raindrop since a circle consumes 78.5% of its
% bounding box.
minArea=min(prod(T.xywh(objPairs(q,:),[3,4]),2))*(1-.785);
overlap(q)=rectint(T.xywh(objPairs(q,1),:),T.xywh(objPairs(q,2),:)) > minArea;
if overlap(q) && all(isvalid(T.obj(objPairs(q,:))))
% highlight the overlapping raindrops, for troubleshooting
% set(T.obj(objPairs(q,:)),'facecolor','m','AmbientStrength',1 )
% Which drop has the smallest width?
[~,minidx]=min(T.xywh(objPairs(q,:),3));
% The smaller drop is absorbed (removed)
T.obj(objPairs(q,minidx),:).delete;
% Elongate the surviving droplet
maxidx=abs(3*(minidx-1)-2); % converts 2 to 1 or 1 to 2;
yd=T.obj(objPairs(q,maxidx)).YData;
ydmu=mean(yd,'all');
ef=1.05; % elongation factor
T.obj(objPairs(q,maxidx)).YData=(ef*(yd-ydmu))+ydmu;
% Update dropSz
T.dropSz(objPairs(q,maxidx))=ef*T.dropSz(objPairs(q,maxidx));
% Make the elongaged drops narrower
xd=T.obj(objPairs(q,maxidx)).XData;
xdmu=mean(xd,'all');
T.obj(objPairs(q,maxidx)).XData=(1/ef*(xd-xdmu))+xdmu;
end
end
% Remove rows of the table that belong to deleted rain drops
T(:,3:end)=[]; % Remove the columns that will be recomputed on next iteration
T(~isvalid(T.obj),:)=[];
end
Animation
Remix Tree