/sci/ - Science & Math » Thread #13963629

75KiB, 875x221, fuckthisequation.png

View Same Google iqdb SauceNAO

Matlab equation

hopelussmatlabier Tue 14 Dec 21:38:45 2021 No.13963629 View Reply Original Report

Quoted By: >>13966706 >>13966885

Hi guys,
I'm wondering if we have any matlab pro's on here that could offer me some help.
(not asking you to do my homework, i've done it i just can't get the expected result and i'm starting to think the assigning lecturer has made a mistake somewhere).

So in the attached picture how would you go about writing it in matlab, for context y' is a 64x64 matrix, as is w1.

I'll post the ways i've attempted so far, hoping there is a guru out there that can point me in the right direction.

hopelussmatlabier

View Same Google iqdb SauceNAO numeroUno.png, 17KiB, 302x374

hopelussmatlabier Tue 14 Dec 2021 21:41:02 No.13963639 Report

Quoted By: >>13963660 >>13965503

Attempt number 1 <<<

hopelussmatlabier

View Same Google iqdb SauceNAO dos+tres.png, 13KiB, 592x116

hopelussmatlabier Tue 14 Dec 2021 21:42:52 No.13963644 Report

Quoted By: >>13965503

Attempt 2 and 3,

They all seem to complete to the same value so i don't think it's the calculation that's wrong.

hopelussmatlabier

hopelussmatlabier Tue 14 Dec 2021 21:45:08 No.13963654 Report

Quoted By:

I did also try using nested loops along the lines of
result = 0
for i = 1:length(w)
for k = 1:length(k)
result = result + sum (sign(y(i,k) * abs(y(i,k) * w(i,k)
end
end

but it had the same result and seemed a bit daft.

Seriously any help would be appreciated.

hopelussmatlabier

hopelussmatlabier Tue 14 Dec 2021 21:47:27 No.13963660 Report

Quoted By:

>>13963639

Strength should be 0.2 in that one.
i'm trying random crap at this point to get it working.

El Arcón

El Arcón Tue 14 Dec 2021 21:49:57 No.13963674 Report

Quoted By: >>13963685

>64x64 matrix to the -0.8 power
nice
>64x64 matrix has its elements written with i but not ij
>very nice

hopelussmatlabier

hopelussmatlabier Tue 14 Dec 2021 21:52:32 No.13963685 Report

Quoted By: >>13966627

>>13963674
i'm not chosing the power unfortunately, the intial image is the spec.

not sure what you mean by your second bit. where did i write with i but not ij?

Anonymous

Anonymous Wed 15 Dec 2021 08:42:28 No.13965503 Report

Quoted By:

>>13963639
looks good
>>13963644
is also valid

Anonymous

Anonymous Wed 15 Dec 2021 17:16:41 No.13966627 Report

Quoted By: >>13967639

>>13963685
maybe try summing the matrix twice, the first sum computes the column sum, and the second one should sum the column

Anonymous

Anonymous Wed 15 Dec 2021 17:44:32 No.13966706 Report

Quoted By: >>13966718 >>13967639

>>13963629
Y = [something];
w = [something];
c = something;
phi = sum(sign(Y) .* abs(Y) .^ (c - 1.0) .* w, 2);

goddamn i hate matlab. why did i ever master it?

Anonymous

Anonymous Wed 15 Dec 2021 17:47:05 No.13966718 Report

Quoted By: >>13967639

>>13966706
i didn't realize that Y and w were matrices. in that case, you can write Y(:) and w(:) to vectorize those arrays and leave everything else intact

Anonymous

Anonymous Wed 15 Dec 2021 18:31:03 No.13966885 Report

Quoted By: >>13967639

>>13963629
x = sum(sign(y).*abs(y).^(c-1).*w1);
detect1=zeros(size(detect1region));
detect1(x>0)=1

hopelussmatlabier

hopelussmatlabier Wed 15 Dec 2021 22:32:47 No.13967639 Report

Quoted By: >>13968117

>>13966627
>>13966706
>>13966718
>>13966885

Thanks all for feeding back.
You were absolutely spot on and the equation was:
result = sum (sign(y) .* abs(y.^(strength-1)) .* w);

Which i had right the whole time.

For context in case anyone is vaguely interested, this was doing 3 levels of DWT decomposition and putting a watermark into the the HL HH and LH subbands. using this equation
y_new = y + 10.*w;
where y_new is the subband new value, y is the subband current value and w is the watermark (again all 64x64)

The thing that had me stuck and cursing matlab for the best part of 20 hours of coding.
The lecturer had switched up the order of the watermarks in the subbands, so instead of putting w1, w2, w3 into LH3, HH3, HL3 respectively. He added them in as w3, w1, w2.

So i was doing everything right just targeting the wrong subbands.

Thanks again for all your comments

Anonymous

Anonymous Thu 16 Dec 2021 01:13:53 No.13968117 Report

Quoted By:

>>13967639
>The thing that had me stuck and cursing matlab for the best part of 20 hours of coding.
matlab is an extremely difficult language to program well. it is sold as being easy, but the language design forces known antipatterns onto programmers making it very painful.
case in point, if you want performance, you need to vectorize as much as possible because loops are slow (not because computers are slow at looping, but because explicit loops in matlab are slower than implicit loops achieved through vectorization). but vectorization is often an optimization, and to do it well you have to design your program around advantageous vectorized operations early.
it's a total cluster fuck for 1000 other reasons

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches