listing 0.9.9 formats |
m →Supported FITS files by Siril: small fix |
||
| Line 1: | Line 1: | ||
=Supported FITS files by [[Siril]]= | =Supported FITS files by [[Siril]]= | ||
In versions 0.9.x, the native image file format of [[Siril]] is unsigned 16-bit FITS (USHORT), one or three layers only. This is the format in which other images are converted to when using the conversion tab, the internal representation of images loaded in Siril and which it processes, the format in which processing results are saved and the format used in FITS sequence export. | In versions 0.9.x, the native image file format of [[Siril]] is unsigned 16-bit FITS (USHORT), one or three layers only, ordered bottom-up. This is the format in which other images are converted to when using the conversion tab, the internal representation of images loaded in Siril and which it processes, the format in which processing results are saved and the format used in FITS sequence export. | ||
==Bitpix: pixel precision== | ==Bitpix: pixel precision== | ||
| Line 8: | Line 8: | ||
* 8-bit integers per channel: the images are converted to 16 bits in memory, no problem | * 8-bit integers per channel: the images are converted to 16 bits in memory, no problem | ||
* 16-bit integers per channel: the signed SHORT images are converted to unsigned short, and no problem either for 15 bits formats. USHORT is in fact [https://heasarc.gsfc.nasa.gov/docs/software/fitsio/c/c_user/node23.html stored as SHORT] with a 2^15 offset. | * 16-bit integers per channel: the signed SHORT images are converted to unsigned short, and no problem either for 15 bits formats. USHORT is in fact [https://heasarc.gsfc.nasa.gov/docs/software/fitsio/c/c_user/node23.html stored as SHORT] with a 2^15 offset. | ||
* 32-bit integers per channel: only the 16 most significant bits are kept, some precision is lost. If in the 32-bit integers contain small data, like values between 0 and 65535, the result will be between 0 and 1 as integer, in other words useless. We may add in the future a way to specify how the conversion is made. | * 32-bit integers per channel: only the 16 most significant bits are kept, equivalent to dividing by 65536, in consequence some precision is lost. If in the 32-bit integers contain small data, like values between 0 and 65535, the result will be between 0 and 1 as integer, in other words useless. We may add in the future a way to specify how the conversion is made. | ||
* 64-bit integers per channel: not currently supported, it would almost be sure to give unusable data. | * 64-bit integers per channel: not currently supported, it would almost be sure to give unusable data. | ||
* 32-bit and 64-bit floating point numbers: they are loaded by multiplying their value with 2^16, so if they are stored with values between 0 and 1 it's usable with some precision loss. If they are stored with values between arbitrary DATAMIN and DATAMAX FITS header keywords values, this will not be usable. | * 32-bit and 64-bit floating point numbers: they are loaded by multiplying their value with 2^16, so if they are stored with values between 0 and 1 it's usable with some precision loss. If they are stored with values between arbitrary DATAMIN and DATAMAX FITS header keywords values, this will not be usable. | ||
Revision as of 00:50, 21 February 2018
Supported FITS files by Siril
In versions 0.9.x, the native image file format of Siril is unsigned 16-bit FITS (USHORT), one or three layers only, ordered bottom-up. This is the format in which other images are converted to when using the conversion tab, the internal representation of images loaded in Siril and which it processes, the format in which processing results are saved and the format used in FITS sequence export.
Bitpix: pixel precision
But Siril can also work with other FITS formats, meaning that it can load, display and process them just like the file format above, but the internal representation, results and exports will still be the native FITS USHORT format, starting with version 0.9.9:
- 8-bit integers per channel: the images are converted to 16 bits in memory, no problem
- 16-bit integers per channel: the signed SHORT images are converted to unsigned short, and no problem either for 15 bits formats. USHORT is in fact stored as SHORT with a 2^15 offset.
- 32-bit integers per channel: only the 16 most significant bits are kept, equivalent to dividing by 65536, in consequence some precision is lost. If in the 32-bit integers contain small data, like values between 0 and 65535, the result will be between 0 and 1 as integer, in other words useless. We may add in the future a way to specify how the conversion is made.
- 64-bit integers per channel: not currently supported, it would almost be sure to give unusable data.
- 32-bit and 64-bit floating point numbers: they are loaded by multiplying their value with 2^16, so if they are stored with values between 0 and 1 it's usable with some precision loss. If they are stored with values between arbitrary DATAMIN and DATAMAX FITS header keywords values, this will not be usable.
Siril 1.0 will switch to a floating point representation of data, providing better precision and support of most formats.
FITS groups
Siril is currently unable to load most FITS images from professional observatories because they contain a lot of metadata and the image data is not registered as the primary content of the file. It is in a subgroup of data, but Siril reads only the first group. Maybe in the future we will handle these formats, but this and the precision of versions 0.9 make Siril unsuitable for professional use.