error in my modified auto-contrast function

Hello guys, I have a project about contrast in image processing and i have to do modified auto-contrast. Here is my code:

function y = modified_autocontrast(a)

a = double(a);

[m, n] = size(a);

a_min = 0;

a_max = 255;

q_low = 0.005;

q_high = 0.005;

n = 256;

count = zeros(1,n);

v = zeros(1,n);

inc = 1;

for p = 0:255

for k = 1:m

for l = 1:n

if(a(k,l) == p)

count(p+1)=count(p+1)+inc;

end

end

end

end

for p = 0:255

if(count(p+1) >= (m*n*q_low))

for c = 0:255

v(c) = count(p+1);

end

end

end

a_low = min(v);

for p = 0:255

if(count(p+1) <= (m*n*(1-q_high)))

for c = 0:255

v(c) = count(p+1);

end

end

end

a_high = max(v);

for i = 1:m

for j = 1:n

if (a <= a_low)

y(i,j) = a_min;

elseif (a_low < a && a < a_high)

y(i, j) = a_min + (a - a_low) * ((a_max - a_min)/(a_high - a_low));

elseif(a >= a_high)

y(i,j) = a_max;

end

end

end

I have an error at this line: v(c) = count(p+1);

How can I fix it? I am new in matlab, help me please

Answer 1

[Dave B]

Your bug: c starts at 0, matlab indexing is one based, you're referring to c(0) but that's not allowed.

Consider: this is a very not MATLAB (and probably slow) approach. Can you do this in a matrix focussed way instead of with loops?

e.g. if the goal of your first loop is to get counts for each value, maybe histcounts would serve you better? Or tabulate (statistics and ML toolbox)? or even don't cast to double and use imhist (image processing toolbox)?

[Ionela Pastinaru]

Thank you a lot! I put v(c + 1) and it works but now it gives me an error at: y(i,j) = a_min; I did the auto contrast function on the same principle, with loops and everything, that's why i want to be like this

[Dave B]

What is the error?

• I don't see how y is initialized in your code.
• a is a matrix and a_low is a scalar...a <= a_low is almost certainly not what you're looking for.

Again the loops are just making your code slow and bug-prone,

data = rand(10);
themin=0; themax=100;
data(data <= 0.2) = themin;
data(data >= 0.8) = themax;
datatoscale(data>0.2 & data <0.8) = themin + (data(data>0.2 & data <0.8) - 0.2) * ((themax-themin)/(0.8 - 0.2));

Or maybe what you're looking for is just:

rescale(data, themin, themax, 'InputMin', .2, 'InputMax', .8);