TODO

* unit tests
* Intel SIMD support for integer and fixed-point transforms
* more wavelets (Haar, Daubechies)
* more data types (fixed-point)
* relicense to some better license
* OpenCV wrapper (C)
* C++ interface
* drop deprecated interface
* remove duplicate code
* tiling (e.g., as in JPEG 2000)
* remove "lib" prefix from source file names
* use size_t, etc. instead of unsigned int, etc.
* ARM support: NEON instructions
* ASVP: upload new data when waiting for preceding operation (use memory bank "D", then copy to bank "A")
* ASVP: number of workers should be adjustable by environment variable like number of threads in OpenMP
* MS Windows support
* code cleanup
* non-dyadic decompositions
* Gabor transform using IIR filters
* functions for type (float, double, int) conversion
* in separable 2-D transforms, remove scaling from horizontal pass and merge it with scaling of vertical pass