Increasing the caclulation speed while using loops

Dear MATLAB experts,

I considered a range for x and y and if x1 and y1 are within a specific range the code must convert x1 and y1 to the mean value of the specific range.

In other words, considering a pixel and if the values of x and y are inside the pixel, I wanted to replace the values with the position of the center of the pixel as shown in the attached file (if x and y are black corss must be replaced with red one)

Code is working but as the data is very big, the calculation speed is slow. Is there any modifications that may lead to increase the speed of calculations?

Thank you in advance.

data = load('data.mat').out1_com;
det_x        = 50;  
det_y        = 50; 
det_ch = 16;
%%
xedges = -(det_x/2):(det_x/det_ch):(det_x/2);
yedges = -(det_y/2):(det_y/det_ch):(det_y/2);

kf   =  data(:,1);
x1   =  data(:,2);
y1   =  data(:,3);
z1   =  data(:,4);
len  =  length(x1);
x11  =  zeros(len,1);
y11  =  zeros(len,1);

for  k=1:length(x1)
    for i = 1:length(xedges)-1
         for j = 1:length(yedges)-1
             idx1 = (xedges(i) <= x1(k)) & ((x1(k)) < xedges(i+1)) & (yedges(j) <= (y1(k))) & ((y1(k)) < yedges(j+1));
             if (idx1 == 1)
                 x11(k)=(xedges(i)+xedges(i+1)) ./ 2;
                 y11(k)=(yedges(j)+yedges(j+1)) ./ 2;
             end
         end
    end  
end

pos=[kf,x11,y11,z1];

[Stephen23]

You could also investigate using HISTCOUNTS2, which will be quite fast.

[Hamid]

@Stephen23 Thank you so much for your comment. The problem is I don't want to count the points in each pixel and want to replace the values with mean value of edges.

[Stephen23]

"The problem is I don't want to count the points in each pixel "

You would use the 4th and 5th outputs (bin indices), not the 1st and 2nd outputs (counts).

Answer 1

[David Goodmanson]

Hi Hamid,

Hi Hamid, I have not speed checked this but it should be faster (the range in your example is different from the drawing).

det_x        = 50;  
det_y        = 50; 
det_ch = 16;
%
xedges = -(det_x/2):(det_x/det_ch):(det_x/2)
yedges = -(det_y/2):(det_y/det_ch):(det_y/2);
xmid = (xedges(1:end-1)+xedges(2:end))/2;
ymid = (yedges(1:end-1)+yedges(2:end))/2;
x = -25+50*rand(1,2000);                  % some data in -25< x,y <25
y = -25+50*rand(1,2000);
xr = round((x+25)*16/50 +1/2);
yr = round((y+25)*16/50 +1/2);
xnew = (xr-1/2)*50/16-25;
ynew = (yr-1/2)*50/16-25;
plot(x,y,'.',xnew,ynew,'o')

[Hamid]

@David Goodmanson

Hi David,

Thank you so much. It was much faster.

Sorry, why you defined the following variables?

xedges = -(det_x/2):(det_x/det_ch):(det_x/2)
yedges = -(det_y/2):(det_y/det_ch):(det_y/2);
xmid = (xedges(1:end-1)+xedges(2:end))/2;
ymid = (yedges(1:end-1)+yedges(2:end))/2;

[David Goodmanson]

Hi Hamid,

I guess you are really asking only about xmid and ymid, since you defined xedges and yedges. I was going to use xmid, ymid to verify that the xnew and ynew arrays were ending up with the right values, but I realized that xnew and ynew were doing the right thing so I never used the check.

[Hamid]

@David Goodmanson

Hi David,

I see. Thanks a lot!