[gdal-dev] Question about JP2Kakadu driver: read on big images may cause "Failed to allocate N byte intermediate decompression buffer for jpeg2000"

[Daniele Romagnoli]

Hi list,
I have a 3 bands 21600x21600 jp2K image and I wish to read it as a
21600x5400 (doing a 4:1 subsampling along Y). [I'm using GDAL 1.5.3]
When doing this read, the DirectRasterIO method gives this error:
"Failed to allocate 1399680000 byte intermediate decompression buffer for
jpeg2000."

Debugging the code I have noticed this:
Firstly, it looks for a proper number of levels to be discarded to reduce
the amount of data to be read.
In this case, being a request of 1:1 along X, no decomposition levels can be
discarded. Then it needs to read the image at its maximum resolution and
then do a subsampling on it.
This step requires to allocate a buffer where to store original data to be
afterwards subsampled. However, depending on the imagesize, this
preallocation (with VSIMalloc) may fail on very big images (my case).

Is there a solution for this problem?

- I have noticed that using GDAL_FORCE_CACHING will use the block based read
method (IReadBlock) which doesn't use that section of code.
- Maybe, when working on big images, splitting the read in several
kdu_stripe_decompressor.pull_stripe calls (with stripeHeights < imageHeight)
on a reduced buffer to be subsampled for each iteration, could be a
solution.

Any suggestions?

Thank you very much.

Regards,
Daniele

-- 
-------------------------------------------------------
Eng. Daniele Romagnoli
Software Engineer

GeoSolutions S.A.S.
Via Carignoni 51
55041 Camaiore (LU)
Italy

phone: +39 0584983027
fax:     +39 0584983027
mob:   +39 328 0559267


http://www.geo-solutions.it

-------------------------------------------------------
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/gdal-dev/attachments/20081125/9cfb3430/attachment-0001.html