Package nom.tam.fits.compression.algorithm.quant
Quantization support for representing floating-point values with integers corresponding to discrete levels. While not a compression alorithm in itself (hence you might also wonder why it's in a package on its own under compression algorithms) quantization is nevertheless commonly used as a pre-sompression (or post-decompression step) with actual algorithms, especially if the algorithms are designed for integer-only data.
Quantization is an inherently lossy process, so it will result in a lossy compression even when paired with a lossless compression algorithm. However, it can significantly improve compression ratios when images have limited dynamic range. For example a 2-byte integer quantization of double-precision values will provide 64k discrete levels at 1/4th of the required storage space -- even before compression is applied.
The only class in here that users would typically interact with is
QuantizeOption
or
CompressedImageHDU.getCompressOption(Class)
to set options after a quantization
algorithm was selected for compressing an image HDU.
-
ClassesClassDescriptionDeprecated.(for internal use) This class sohuld have visibility reduced to the package levelQuantization options when they are part of the compression scheme.(for internal use) Qunatization step processor as part of compression.TODO this is done very inefficient and should be refactored!A standard fixed random sequence to use for portable and reversible dither implementations.