search

imglib / imglib2

A generic next-generation Java library for image processing
http://imglib2.net/
Other
302 stars 90 forks source link

Consider moving BigComplex, PreciseFixedFloatType, etc. to ImgLib2 #107

Open awalter17 opened 9 years ago

awalter17 commented 9 years ago

Hello!

I noticed today that BigComplex, PreciseFixedComplexFloatType, PreciseFixedFloatType, and UnboundedIntegerType are all in ImageJ-Common. However, I think these classes would fit better in ImgLib2 with all the other types like IntType, UnsignedByteType, etc.

Please let me know your thoughts on this.

tpietzsch commented 8 years ago

@awalter17 I think they should only be in core if they are really used anywhere or we expect them to be used. The only application scenarios I see is using them as intermediate results in calculations. You would probably not make images of them. They are not 'NativeType's, so the only option would be making 'ListImg', so limited to quite small images.

There is really no reason to put stuff into core that is just "proof-of-principle". Along these lines, we should think about removing BasePairBitType and BasePairCharType. I think, they are not really useful except for demonstration purposes. But not sure. If they are in use, they should be remain core of course. @StephanPreibisch @axtimwalde Opinions?

@awalter17 @ctrueden Do you know whether the ImageJ-Common types are used?

ctrueden commented 8 years ago

I think arbitrary precision maths are very important for science, and will grow increasingly more important as computers continue to grow in their capabilities. So my vote is for ImgLib2 to provide an out-of-the-box way of doing them.

so limited to quite small images.

Not necessarily, if we support cell-based caching for them in the future. Then it's just a matter of computation time.

Do you know whether the ImageJ-Common types are used?

Well, they would start to be used if we added them to ImageJ's Image ▶ Type menu.