BioFormat for matlab and svs images

Dear all

does anyone know how can i open an svs image by using bioformat?

My images have 4 planes:

• image at the maximum resolution
• image at the mid resolution
• image at 1/8 resolution (or even less)
• image with the original resolution bar.

When trying to open an image to get the second plane, I use the following commands:

r=bfGetReader('C:\prova.svs'); 

when issued, this command raises this error:

log4j:WARN No appenders could be found for logger (loci.formats.ClassList).
log4j:WARN Please initialize the log4j system properly.

if I further proceed with the bfGetPlane instruction:

img=bfGetPlane(r,2);

I get the following error:

Error using bfGetPlane (line 78)
Java exception occurred:
java.lang.IllegalArgumentException: Array size too large: 89488 x 74009 x 1

	at loci.common.DataTools.safeMultiply32(DataTools.java:949)

	at loci.common.DataTools.allocate(DataTools.java:922)

	at loci.formats.ChannelSeparator.openBytes(ChannelSeparator.java:158)

Please, help me!

[Walter Roberson]

That image appears to be about 6.8 gigabytes, assuming that the components are uint8 .

You would have to go into Preferences to change Java Memory to permit at least that much memory for Java. You would typically need to restart MATLAB once after increasing Java memory.

[Elena Casiraghi]

Thanks Walter, I already did that. Then I should suppose there is no solution.

[Walter Roberson]

The routine name loci.common.DataTools.safeMultiply32 hints that possibly it might need 32 bits per component, or 4 times 6.8 gigabytes.

Or perhaps it hints that it uses 32 bit addresses and so might be restricted to 2 or 4 gigabytes.

[Walter Roberson]

I see from the source code at https://github.com/openmicroscopy/bioformats/blob/develop/components/formats-gpl/matlab/bfGetPlane.m that just before the openBytes call, that it calls

pixelType = r.getPixelType();
bpp = javaMethod('getBytesPerPixel', 'loci.formats.FormatTools', pixelType);
fp = javaMethod('isFloatingPoint', 'loci.formats.FormatTools', pixelType);
sgn = javaMethod('isSigned', 'loci.formats.FormatTools', pixelType);
little = r.isLittleEndian();

It would be interesting to see the result of bpp and fp, which you could get by putting a breakpoint in at the openBytes call and examining the variables. This would help figure out what array size was needed.

It looks to me as if it might be possible to use bfGetPlane to read in portions of the image at a time. Even if you have to piece the parts together again, you would be doing that in main RAM instead of in java memory, so this approach could help get around java limits.

[Elena Casiraghi]

bpp=1 fp logical 0 sgn logical 0 I used also bfGetPlane but it gave the same error