Login  Register

Re: jpg2000 and mxrs in imagej

Posted by Herbie on Dec 06, 2015; 5:28pm
URL: http://imagej.273.s1.nabble.com/jpg2000-and-mxrs-in-imagej-tp5015148p5015151.html

Filip,

I found your explanations quite helpful and I hope that those on the
list who may help further, also get a better impression of the kind of
images you are dealing with.

If you can separately access the 4096x4096 images, my recommendation is
to do the tiling in Fiji which more or less implies to rely on TIF as
the file format.

HTH

Herbie

:::::::::::::::::::::::::::::::::::::::::::
Am 06.12.15 um 18:22 schrieb Filip Bochner:

> Herbie,
>
> Of course you're right. I couldn't see what is the bit depth in
> properties of the file. I would also guess it's 24bit. Apparently
> Image Viewer displayed it incorrectly. And I trusted it
> unfortunately.
>
> Yes, we keep talking about Fiji :)
>
> The slide scanner' name's Panoramic Midi. It saves the files in .mrxs
> format. Each .mrxs is accompanied with a folder of .dat files. In
> principle it's a tiled image that can be viewed by free Panoramic
> Viewer (that doesn't have a batch export and it's poorly written so
> the conversion of single ROI takes forever and I have tens of slides
> with tens of ROIs and will have even more). A single scan gives
> 4096x4096 resolution images, but there is an array of them, which
> after export gives a huge matrix.
>
> I don't think it's a size (gb) problem, since I opened much larger
> files in Fiji. It could be an array-size problem. It might not be
> even related to the format, but I'm not sure whether .tiff of similar
> size would give the same error (or crash).
>
> Filip
>
> Wysłane z iPada
>
> Dnia 06.12.2015 o godz. 17:26 Herbie <[hidden email]> napisał(a):
>
>> Filip,
>>
>> evidently, RGB 0 bit doesn't make sense -- no?
>>
>> Usually, RGB means 24bit per pixel which results in about 10.5
>> GByte for your image. I don't know why ImageJ2 can't deal with 10.5
>> GByte images, but I know that the old data structures of Java,
>> hence ImageJ, are can't.
>>
>> "[...] the program cannot load the files" I guess with "program"
>> you mean "Fiji".
>>
>> BTW, which _slide_ scanner provides "33280 x 105216" pixels per
>> image? What kind of slides (format?) do have such spatial
>> resolution?
>>
>> Best
>>
>> Herbie
>>
>> :::::::::::::::::::::::::::::::::::::::::::
>>> Am 06.12.15 um 16:04 schrieb Filip Bochner: Hi Herbie,
>>>
>>> "x 1" means that this is just one image. If it was a stack with
>>> let's say 13 images it would say "x 13".
>>>
>>> NIS-elements viewer (which just opens the images) shows RGB 0
>>> bit.
>>>
>>> Filip
>>>
>>> -----Original Message----- From: ImageJ Interest Group
>>> [mailto:[hidden email]] On Behalf Of Herbie Sent: Sunday,
>>> December 06, 2015 4:46 PM To: [hidden email] Subject: Re:
>>> jpg2000 and mxrs in imagej
>>>
>>> Filip,
>>>
>>> I've no experience with ImageJ2 data structures and don't know
>>> about their limits, but what does "x 1" mean in "33280 x 105216 x
>>> 1"? Does it indicate a single image, i.e. no stack, or does it
>>> stand for a binary-valued image?
>>>
>>> If it is not a binary-valued image, what is its bit-depth?
>>>
>>> Best
>>>
>>> Herbie
>>>
>>> ::::::::::::::::::::::::::::::::::::::::::: Am 06.12.15 um 14:56
>>> schrieb Filip Bochner:
>>>> Hi,
>>>>
>>>> I'm struggling with export of very large .mrxs (mirax) files.
>>>> These are scans from a slide scanner that I intend to analyze
>>>> further in ImageJ.
>>>>
>>>> I found a tool called JP2 WSI converter. It seems really good,
>>>> because it can batch-convert huge mrxs files into jpg2000.
>>>>
>>>> The problem is that even if ImageJ2 data structures and SCIFIO
>>>> are enabled, the program cannot load the files.
>>>>
>>>> Before I enabled them I was getting a following message:
>>>> java.lang.IllegalArgumentException: Array size too large: 33280
>>>> x 105216 x 1
>>>>
>>>> After enabling ImageJ2 data structures and SCIFIO the memory
>>>> gets flooded and the software gets stuck (work on a workstation
>>>> with 6 cores, 6 threads and 64 GB of memory).
>>>>
>>>> Thanks for help!
>>>>
>>>> Filip
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html