Download Latest Version pfstools-2.2.0.tgz (614.8 kB)
Email in envelope

Get an email when there's a new version of pfstools for HDR images

Name Modified Size InfoDownloads / Week
Parent folder
README.txt 2011-03-11 2.1 kB
pfscalibration-1.5.tar.gz 2011-03-11 341.8 kB
Totals: 2 Items   344.0 kB 0
This release of pfscalibration fixes many issues with handling RAW
(more than 8-bit) images. Robertson's algorithm should produce less
banding. Mitsunaga and Nayar's algorithm is still broken. Two new
scripts have been added: pfsinme - to skip generating an intermediate
hdrgen file; and pfsplotresponse - for plotting camera response
functions.

----
2011-03-04 1.5 Rafal Mantiuk
	* added pfsplotresponse script for plotting camera response curves
	* fixed the right range of values is used in Mitsunaga's algorithm. However this algoritm still seems to be broken. Use Robertson instead.
	* updated: manual page for pfshdrcalibrate

2011-02-15 Ivo Ihrke
	* fixed multiplier bug for raw data > 8bit
	  - use --bpp flag if you want to take adavantage of that
	    the program will assume 8 bit by default, even for raw files
	    like (.cr2 for Canon which, e.g., stores 14bit/pixel for
	    a 5D Mark II)
	* reorderd computations such that an HDR image is computed also
	  in the case of response curve estimation
	* Gaussian replaced by robust Gaussion (responses.cpp)
	  this fixes the saturation problems, the hacks to deal with that
	  have been removed (robertson.c applyResponse)
	* fixed infinite loop due to incorrect convergence estimate in
	  robertson02.cpp getResponse function
	* !! changed behavior of -g flag (Gaussian kernel size)
	     it is now a value relative to the [0-1] range
	     default value changed to 0.2

	* changed the weight function from Gaussian to composite
	  goals: - suppress noise on the dark end
	       - use bright intensity levels in a better way (higher SNR of these pixels)
	       - prevent use of bright saturated pixels
	       - do not zero bright saturated pixels

	  - noise behavior is much better than simple robust Gaussian,
	    the saturated pixels problem and color banding
	    do occur in a much less pronounced way
	* changed file format of response curve file, it now uses
	  scientific notation for floats (this improves precision,
	  especially for the weight function), the files can still be read
	  by older versions of pfstools

2008-10-02
	* added: pfsinme command
Source: README.txt, updated 2011-03-11