Reading uint8 from raw data takes much more RAM then expected

I am trying to read some MB of data from big files of several GB. I am using the follwing code:

fid=fopen(filename,'r');
fseek(fid,someStartValue,'bof');
raw = fread(fid, [3 ceil(numOfPixels/2)*numOfFrames],'uint8','b');

number of uint8 are

3*ceil(numOfPixels/2)*numOfFrames=184688640

so this is also the value in bytes, corresponding to about 185MB. But when calling the fread line, the taken RAM increases by 1.6 GB.

Any ideas?

Running the fread line mutiple times in a row gives the peaks shown in the attached image. I also don't understand, why there is more RAM available after the peak compared to before the peak. The image shows three calls of the line

[Askic V]

Hello Jonas, did you chek this:

https://www.mathworks.com/help/matlab/matlab_prog/strategies-for-efficient-use-of-memory.html

[Jonas]

thanks for the link. Guessing fread internally stores those values as double first, the size factor is 8 between uint8 and double which would equal about 1.5 GB. sounds like that could be the problem. tomorrow i will try

raw = fread(fid, [3 ceil(numOfPixels/2)*numOfFrames],'uint8=>uint8','b');

[Les Beckham]

That should fix your issue. If you specify the precision argument as 'uint8' as your initial post shows, it will take the uint8 source data and put it into doubles. Using 'uint8=>uint8' will keep the data as uint8, reducing the memory usage by a factor of 8. As Image Analyst suggests in his answer, use the whos command to check the data type and memory size of your raw array after you read it in .

[Jonas]

issue solved, thanks.

Answer 1

[Image Analyst]

Put the memory command before and after your code. What do you see? Are you sure you're closing the file?

memory
fid=fopen(filename,'r');
fseek(fid,someStartValue,'bof');
raw = fread(fid, [3 ceil(numOfPixels/2)*numOfFrames],'uint8','b');
fclose(fid);
whos raw
memory