feat(arw): Adding automatic conversion for Sony ARW Camera formats

[luisgreen]

Automatic Sony ARW raw file conversion & metadata printing

Summary

Adds automatic detection, metadata extraction, and conversion of Sony ARW (.ARW) raw camera files to 16-bit TIFF before denoising. Users can now point denoise_image.py directly at ARW files without manual preprocessing.

• Detection: The function convert_arw_if_needed() checks the input path's extension (case-insensitive .ARW)
• Metadata: On ARW detection, prints camera make/model (via exiftool), resolution (from rawpy), ISO, shutter speed, aperture, and focal length (via rawpy.RawPy.other)
• Conversion: Uses rawpy.postprocess(use_camera_wb=True, output_bps=16) to produce a 16-bit RGB array, then writes it as TIFF via OpenCV
• Temp file: Created via tempfile.mkstemp() in the same directory as the ARW file; cleaned up in a finally block after processing
• Transparency: Non-ARW inputs (TIFF, JPEG, PNG) pass through unchanged; output naming uses the original ARW filename, not the temp filename
• Error handling: Graceful exit with instructions if rawpy is not installed; error message on corrupt files
• Example output

Detected ARW raw file: DSC02375.ARW
  Camera: SONY ILCE-7RM3
  Resolution: 7968 x 5320
  ISO: 6400.0
  Shutter: 1/500 s
  Aperture: f/4.0
  Focal length: 200 mm
  Converted to temporary TIFF: /path/to/tmpXXXX.tiff

[luisgreen]

BTW... Came here just by chance... and ended up being exactly what I needed. Thanks for your efforts!

[trougnouf]

Thanks for the PR :-)

I'm not sure this is the best use-case :s Did you try https://github.com/trougnouf/rawnind_jddc ? It's made specifically for raw images, it denoises the raw files and outputs a virtually raw tiff file (linear RGB with just the demosaic applied) that can be developed in any way (e.g. with darktable), and it should provide better denoising performance.

If rawnind-denoise is not the experience you are looking for and you truly benefit more from this use case (getting a ready-to-publish denoised jpeg from a raw file) then we can merge this (although there is no reason to limit it to Sony ARW files since libraw/RawPy handles any/most raw image formats seamlessly), otherwise I would rather point users in the direction of RawNIND.

[luisgreen]

Thanks for the PR :-)

I'm not sure this is the best use-case :s Did you try https://github.com/trougnouf/rawnind_jddc ? It's made specifically for raw images, it denoises the raw files and outputs a virtually raw tiff file (linear RGB with just the demosaic applied) that can be developed in any way (e.g. with darktable), and it should provide better denoising performance.

If rawnind-denoise is not the experience you are looking for and you truly benefit more from this use case (getting a ready-to-publish denoised jpeg from a raw file) then we can merge this (although there is no reason to limit it to Sony ARW files since libraw/RawPy handles any/most raw image formats seamlessly), otherwise I would rather point users in the direction of RawNIND.

Thank you for your response. let me take a look at it!