gphoto2 + better exposure estimation

Additions:

• Programmatically capture images with gphoto2 by connecting a DSLR
• Improved exposure estimation with MST-reduction and Tikhonov penalty; quicker method agnostic to resolution
• Merge returns an unsaturated mask for post-processing without color artifacts

Robust exposure estimation

• Includes exposure estimation of Cerman and Hlavac
• Better scatter plots to check linearity
• Iterative outlier removal for robust exposure estimation
• Align RAW bayer images

Automatic exposures

• Updated noise simulation interface
• Automatic exposure estimation to account or illumination changes
• Added tests for common functions

Simulate noise

Additional functionalities:

• Simulate calibrated cameras using simple noise models
• Write FP16 HDR images
• Merge non-demosaiced PNG images

Updated colorspace and alignment

This update has the following additions and fixes:

• Merge before demosaicing mode now supports more colorspaces (sRGB, Adobe)
• Feature matching for homography done on scaled images
• Better logs and warnings for alignment

Minor bug fixes

Summary of changes:

• Better 16-bit png support
• Clip negative pixels after HDR merging
• Improved logs
• Continue merging even with failed alignment

Minor bug fixes and improved logs

Following bugs were addressed:

• OpenCV SIFT failed when too many keypoints were detected
• Rawpy processing flipped images taken in portrait mode, leading to incorrect height and width
• Tested on Canon cameras and changed white point to percentage-based value

Additionally, some logging messages were addressed and HDRutils now supports LDR file formats as well.