Measuring average intensity of pixels for multiple images.

I am trying to read multiple similar images located in a folder and find the average intensity of pixels of the images. Following is the code I wrote initially which works fine:

Location = 'D:\\dummy location';

B = dir(fullfile(sprintf(Location),sprintf('BLACK'), '*.tif'));
B_0 = numel(B);
for k = 1:B_0
    F = fullfile(sprintf(Location),sprintf('BLACK'),B(k).name);
    I{k} = imread(F);
end
B1 = cat(3,I{:});
black = mean(B1,3);

However, upon getting suggested to use vectorization to make the process faster, I wrote the following code:

Location = 'D:\\dummy location';

B = dir(fullfile(sprintf(Location),sprintf('BLACK'), '*.tif'));
B_0 = numel(B);
k = 1:B_0;
F = fullfile(sprintf(Location),sprintf('BLACK'),B(k).name);
I{k} = imread(F);

B1 = cat(3,I{:});
black = mean(B1,3);

I am getting the following error:

Error using dir

Pathname 'D:\dummy location\BLACK\1.tif\10.tif\100.tif\101.tif\102.tif\103.tif\104.tif\105.tif\106.tif\107.tif\108.tif\109.tif\11.tif\110.tif\111.tif\112.tif\113.tif\114.tif\115.tif\116.tif\117.tif\118.tif\119.tif\12.tif\120.tif\121.tif\122.tif\123.tif\124.tif\125.tif\126.tif\127.tif\128.tif\129.tif\13.tif\130.tif\131.tif\132.tif\133.tif\134.tif\135.tif\136.tif\137.tif\138.tif\139.tif\14.tif\140.tif\141.tif\142.tif\143.tif\144.tif\145.tif\146.tif\147.tif\148.tif\149.tif\15.tif\150.tif\151.tif\152.tif\153.tif\154.tif\155.tif\156.tif\157.tif\158.tif\159.tif\16.tif\160.tif\161.tif\162.tif\163.tif\164.tif\165.tif\166.tif\167.tif\168.tif\169.tif\17.tif\170.tif\171.tif\172.tif\173.tif\174.tif\175.tif\176.tif\177.tif\178.tif\179.tif\18.tif\180.tif\181.tif\182.tif\183.tif\184.tif\185.tif\186.tif\187.tif\188.tif\189.tif\19.tif\190.tif\191.tif\192.tif\193.tif\194.tif\195.tif\196.tif\197.tif\198.tif\199.tif\2.tif\20.tif\200.tif\201.tif\202.tif\203.tif\204.tif\205.tif\206.tif\207.tif\208.tif\209.tif\21.tif\210.tif\211.tif\212.tif\213.tif\214.tif\215.tif\216.tif\217.tif\218.tif\219.tif\22.tif\220.tif\221.tif\222.tif\223.tif\224.tif\225.tif\226.tif\227.tif\228.tif\229.tif\23.tif\230.tif\231.tif\232.tif\233.tif\234.tif\235.tif\236.tif\237.tif\238.tif\239.tif\24.tif\240.tif\241.tif\242.tif\243.tif\244.tif\245.tif\246.tif\247.tif\248.tif\249.tif\25.tif\250.tif\251.tif\252.tif\253.tif\254.tif\255.tif\256.tif\257.tif\258.tif\259.tif\26.tif\260.tif\261.tif\262.tif\263.tif\264.tif\265.tif\266.tif\267.tif\268.tif\269.tif\27.tif\270.tif\271.tif\272.tif\273.tif\274.tif\275.tif\276.tif\277.tif\278.tif\279.tif\28.tif\280.tif\281.tif\282.tif\283.tif\284.tif\285.tif\286.tif\287.tif\288.tif\289.tif\29.tif\290.tif\291.tif\292.tif\293.tif\294.tif\295.tif\296.tif\297.tif\298.tif\299.tif\3.tif\30.tif\300.tif\301.tif\302.tif\303.tif\304.tif\305.tif\306.tif\307.tif\308.tif\309.tif\31.tif\310.tif\311.tif\312.tif\313.tif\314.tif\315.tif\316.tif\317.tif\318.tif\319.tif\32.tif\320.tif\321.tif\322.tif\323.tif\324.tif\325.tif\326.tif\327.tif\328.tif\329.tif\33.tif\330.tif\331.tif\332.tif\333.tif\334.tif\335.tif\336.tif\337.tif\338.tif\339.tif\34.tif\340.tif\341.tif\342.tif\343.tif\344.tif\345.tif\346.tif\347.tif\348.tif\349.tif\35.tif\350.tif\351.tif\352.tif\353.tif\354.tif\355.tif\356.tif\357.tif\358.tif\359.tif\36.tif\360.tif\361.tif\362.tif\363.tif\364.tif\365.tif\366.tif\367.tif\368.tif\369.tif\37.tif\370.tif\371.tif\372.tif\373.tif\374.tif\375.tif\376.tif\377.tif\378.tif\379.tif\38.tif\380.tif\381.tif\382.tif\383.tif\384.tif\385.tif\386.tif\387.tif\388.tif\389.tif\39.tif\390.tif\391.tif\392.tif\393.tif\394.tif\395.tif\396.tif\397.tif\398.tif\399.tif\4.tif\40.tif\400.tif\401.tif\402.tif\403.tif\404.tif\405.tif\406.tif\407.tif\408.tif\409.tif\41.tif\410.tif\411.tif\412.tif\413.tif\414.tif\415.tif\416.tif\417.tif\418.tif\419.tif\42.tif\420.tif\421.tif\422.tif\423.tif\424.tif\425.tif\426.tif\427.tif\428.tif\429.tif\43.tif\430.tif\431.tif\432.tif\433.tif\434.tif\435.tif\436.tif\437.tif\438.tif\439.tif\44.tif\440.tif\441.tif\442.tif\443.tif\444.tif\445.tif\446.tif\447.tif\448.tif\449.tif\45.tif\450.tif\451.tif\452.tif\453.tif\454.tif\455.tif\456.tif\457.tif\458.tif\459.tif\46.tif\460.tif\461.tif\462.tif\463.tif\464.tif\465.tif\466.tif\467.tif\468.tif\469.tif\47.tif\470.tif\471.tif\472.tif\473.tif\474.tif\475.tif\476.tif\477.tif\478.tif\479.tif\48.tif\480.tif\481.tif\482.tif\483.tif\484.tif\485.tif\486.tif\487.tif\488.tif\489.tif\49.tif\490.tif\491.tif\492.tif\493.tif\494.tif\495.tif\496.tif\497.tif\498.tif\499.tif\5.tif\50.tif\500.tif\51.tif\52.tif\53.tif\54.tif\55.tif\56.tif\57.tif\58.tif\59.tif\6.tif\60.tif\61.tif\62.tif\63.tif\64.tif\65.tif\66.tif\67.tif\68.tif\69.tif\7.tif\70.tif\71.tif\72.tif\73.tif\74.tif\75.tif\76.tif\77.tif\78.tif\79.tif\8.tif\80.tif\81.tif\82.tif\83.tif\84.tif\85.tif\86.tif\87.tif\88.tif\89.tif\9.tif\90.tif\91.tif\92.tif\93.tif\94.tif\95.tif\96.tif\97.tif\98.tif\99.tif'

is too long.

Error in imread>get_full_filename (line 560)

if ~isempty(dir(filename))

Error in imread (line 371)

fullname = get_full_filename(filename);

Error in avg_new (line 12)

I{k} = imread(F);

Can anyone please help me correcting this error to get the same result as my first code?

[DGM]

Your first code is fine. There isn't a way to feed imread() a wildcard or an array of filenames, so vectorization doesn't make sense here.

I'm not sure why you're doing it that particular way. If you're assuming that all the images are single-channel, and they have the same page geometry, then you can do this a number of ways. Consider the three examples

In this example, all the images are loaded into a cell array and averaged

% i'm using my own test files (1521 files, 286x384 px)
Location = 'sources/faces';
B = dir(fullfile(Location,'*.pgm'));
B_0 = numel(B);

tic
I = cell(B_0,1); % preallocate
for k = 1:B_0
    F = fullfile(Location,B(k).name);
    I{k} = imread(F);
end

meanvalues = cast(mean(cat(3,I{:}),3),class(I{1})); % average and recast
toc

In this example, all the images are loaded as a volumetric image and then averaged all at once.

% i'm using my own test files (1521 files, 286x384 px)
Location = 'sources/faces';
B = dir(fullfile(Location,'*.pgm'));
B_0 = numel(B);

tic
for k = 1:B_0
    F = fullfile(Location,B(k).name);
    I = imread(F);
    
    if k == 1
        meanvalues = zeros([size(I,1:2),B_0],class(I));
    end
    
    meanvalues(:,:,k) = I; % store all images as pages in a volume
end

meanvalues = cast(mean(meanvalues,3),class(I)); % calculate pagewise mean of volume
toc

In this example, only one image is loaded at a time, incrementally accumulating a sum prior to division.

% i'm using my own test files (1521 files, 286x384 px)
Location = 'sources/faces';
B = dir(fullfile(Location,'*.pgm'));
B_0 = numel(B);

tic
for k = 1:B_0
    F = fullfile(Location,B(k).name);
    I = imread(F);
    
    if k == 1
        meanvalues = double(I); % allocate
    else
        meanvalues = meanvalues + double(I); % accumulate
    end
end

meanvalues = cast(meanvalues/B_0,class(I)); % divide and recast
toc

The first examples might seem faster because all the arithmetic is done only once. For a small set of images, it might be faster, but as the total volume of data increases, the time spent allocating memory will slow it down. The potential use of cell arrays may or may not improve the time-cost of that memory allocation in certain cases.

In my own test, the first two examples are slightly slower than the incremental method in the second example, but that depends on the total size of the data, your amount of memory, etc. The cell array method at least succinct.

[Mukul]

Thank you for the three methods. I appreciate your help.

Answer 1

[Image Analyst]

There is absolutely no need to store all your images in a cell array or as a 3-D image.

See the FAQ for code on how to process a sequence of images:

Process a sequence of files

Also see this FAQ entry on cell arrays: What is a cell array

[Mukul]

Thank you, these links are highly useful!