imresize using NaN and parfor causes memory error

I've used imresize (using the NaN option) in a parfor and I've got a memory error. However, without the NaN option (see case 2 below - namely by calculating the target dimension in advance) it worked perfectly

e.g.

case 1: I2=imresize(I,[256 NaN]);

case 2: I2=imresize(I,[256 256]);

• case 1 or case 2 + for, OK
• case 2 + parfor, OK
• case 1 + parfor, memory error

Do you have any idea why this is happening?

I thank you in advance

Vassilis

[Edric Ellis]

I tried the following in R2014b, and it worked fine. Could you post a simple self-contained example that fails.

imdata = imread('ngc6543a.jpg');
parfor idx = 1:1
  i2{idx} = imresize(imdata, [256, NaN]);
end

[Vassilios]

sure, this is an example (it is useless because it scales the same image but it can show the error)

>> I=imread('cameraman.tif');
>> for i=1:10, I2{i}=imresize(I,[256 NaN]); end
>> clear I2
>> parfor i=1:10, I2{i}=imresize(I,[256 NaN]); end
Out of memory. Type HELP MEMORY for your options.

while this

parfor i=1:10, I2{i}=imresize(I,[256 128]); end

works perfectly.

Note that I'm using R2014a

[Edric Ellis]

I just tried your reproduction in R2014a on WIN64 and GLNXA64, and it worked fine - what OS are you using? Does it make any difference if you open a smaller pool? I.e. try

parpool('local', 1)

When you get the error, could you try:

E = MException.last

and if E is a ParallelException (which means that the out-of-memory occurred on the workers rather than on the client), also try

getReport(E.remotecause{1})