RE: i1Profiler scanner module is very picky about file formats
RE: i1Profiler scanner module is very picky about file formats
- Subject: RE: i1Profiler scanner module is very picky about file formats
- From: Ethan Hansen <email@hidden>
- Date: Wed, 11 Feb 2015 20:53:33 -0800
- Organization: Dry Creek Photo
From: Roger Breton
> This is for X-Rite... Sorry for posting here :( But I'm in i1Profiler 1.5.6 demo on Windows 8.1 and I am having one hell of a time to get a Kodak Transparent 4.5 TIFF 300dpi image to get the fiducial marks on? I tried many things to no avail. The last time I had difficulty with this error, increasing the canvas size in Photoshop did the trick. But, after trying twice to increase the canvas size, i1Profiler crashed on me with :
> OpenCV GUI Error Handler
> Insufficient memory (Out of memory)
> In function cvAlloc, C:\Users\NYadav\Downloads\Openvc-1.0\cxcore\src\cxalloc.cpp(111)
> I guess, the only thing left to try is to press the Abort button :(
Well, that's an error I've seen before. Not in i1Profiler but in some of our old code that used OpenCV. The problem often is not that the application actually ran out of memory, but that memory was corrupted and overwritten. Memory leaks are all too prevalent in i1Profiler. Do too many tasks in a row and it's crash time.
Had you run anything else in i1Profiler before trying to load your large canvas image? If not, try it from a clean start.
----------------------
Notes to X-Rite:
- OpenCV happily consumes gobs of memory. A pointer to an image object *must* be released before it is reassigned or reused. For example, when you allocate IplImage *pImg, make sure to call cvReleaseImage(&pImg).
- Even with careful programming, the C interface to OpenCV is prone to memory leaks. Using the C++ interface and thereby replacing IplImage with cv:Mat makes your life considerably easier.
- That said, track memory usage before allocating space for new images. Also, minimize the number of images and vectors to images stored at once.
- Finally, if resampling is being performed, we found a quirk with cvResize. Using bicubic interpolation (CV_INTER_CUBIC) led to minor memory leaks while other interpolation algorithms did not. We killed the project shortly thereafter, so I can't point a finger at bicubic interpolation with any certainty. In any event, if scanner images are being resized, Nearest Neighbor (CV_INTER_NN) should be used to preserve color accuracy.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden