In general images are almost never a power of two.  In various parts of my
application this holds true for the data I would like to dump to hdf.  I
presently explicitly set the line step bytes to be columns * pixelsize, but
in general this is not good because many optimized functions expect each
row to start on a 16 or 32 byte boundary.   Also, please keep in mind
images can be uint8, uint16, uint32, floats...


-Jason


On Thu, Sep 19, 2013 at 8:10 AM, Werner Benger <[email protected]> wrote:

> **
> Would it be appropriate for your application to use a chunked dataset such
> that each chunk is sized to e.g. a power of 2 ?
>
>              Werner
>
>
> On Thu, 19 Sep 2013 09:36:25 -0500, Jason Newton <[email protected]> wrote:
>
> Hi,
>
> Resurrecting this thread:
>
> http://mail.lists.hdfgroup.org/pipermail/hdf-forum_lists.hdfgroup.org/2012-November/006206.html
>
> Although the original author mentions bits, he means bytes.  It's a
> regular thing that images have a row of bytes referred to as a line to
> ensure alignment of each pixel row.  This usually is padded with up to 15
> bytes.  As the original author indicates numerous libraries do this.  From
> image IO (libtiff: scanline) to image processing libraries like  opencv/ipp
> (step).
>
> However HDF only likes packed arrays and doesn't really deal with this
> notion of storage.  So how can we make it deal with it not in terms of the
> data space but in terms of something some detail to this  particular memory
> buffer? Copying images to packed format is an ugly and non-optimal
> solution;  a bunch of write calls isn't great either.
>
> Regards,
> -Jason
>
>
>
>
> --
> ___________________________________________________________________________
> Dr. Werner Benger Visualization Research
> Center for Computation & Technology at Louisiana State University (CCT/LSU)
> 2019 Digital Media Center, Baton Rouge, Louisiana 70803
> Tel.: +1 225 578 4809 Fax.: +1 225 578-5362
>
_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]
http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org

Reply via email to