File: ChangeLog

package info (click to toggle)
pfstmo 1.4-1
  • links: PTS, VCS
  • area: main
  • in suites: jessie, jessie-kfreebsd, squeeze, wheezy
  • size: 2,108 kB
  • ctags: 551
  • sloc: sh: 9,063; cpp: 5,363; ansic: 872; makefile: 139
file content (208 lines) | stat: -rw-r--r-- 6,606 bytes parent folder | download
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
23/02/2009 PFSTMO 1.4
	* mantiuk08: added color correction method from (http://zgk.wi.ps.pl/color_correction/)
	* mantiuk08: fixed dependency on libgsl (thanks to Niels)
	* durand02: added color correction method from (http://zgk.wi.ps.pl/color_correction/)
	
04/11/2008
	* all: variables with underscore replaced with "n<variable name>"
	as they caused compilation problems under cygwin
	
21/09/2008
	* mantiuk08: Refactored for exception safety

21/09/2008

	* build: Removed bogus flag -march=nocona; added CPU detection and
	--disable-portable option for enabling non-portable optimizations.
	Tunes for local CPU by default, can be disabled using
	--without-gcc-arch

21/09/2008

	* pfstmo.h: Templatized Array2D; added copy constructor,
	assignment operator, and function `allocate' for re-initializing
	array after construction

09/09/2008 PFSTMO 1.3.2

	* durand02: removed downsampling option - introduced large errors
	into the results
	
08/09/2008

	* durand02: division by zero bug fixed (fastbilateral)
	
	* durand02: range boundary bug fixed (fastbilateral)

	* durand02: added progress reporting (fastbilateral)

	* durand02: performance imrovements in fastbilateral
	
	* mantiuk08: found the cause for the uninitialized value in the cqpminimizer.
	Initialization moved to more a suitable code location
	
05/09/2008

	* durand02: disabled fast_bilateral after finding some bugs. A
	slow bilateral filter will be used until bugs are fixed.

04/09/2008

	* all: removed dependency on pfs::Array2D[Impl], replaced by new
	class pfstmo::Array2D.  Apart from getting rid of the dependency,
	this also speeds thing up considerably because pfstmo::Array2D is
	non-virtual and thus can be inlined.
	
03/09/2008

	* durand02: added missing delete in fastBilateralFilter. All TMOs
	are now valgrind memcheck-clean.

03/09/2008

	* mantiuk08: fixed uninitialized value in gsl_cqpminimizer_alloc.

02/09/2008

	* fattal02: replaced delete with correct delete[] operators in PDE
	solver.

27/08/2008

	* fixed: added missing FFTW3F_CFLAGS for durand02 operator 
	
	* pattanaik00: Fixed a bug that would cause every frame to be
	treated as the first frame, never invoking
	VisualAdaptationModel::calculateAdaptation.
	
27/08/2008

	* configure.ac: Added missing likely/unlikely macros

26/08/2008 PFSTMO 1.3.1

	* mantiuk06: Ed Brambley's bug-fix and convergence patch:

	 - There was a bug in the upsampling code that could cause an
	   out-of- bound read past the end of allocated memory.  This was
	   occasionally causing NaNs to creep in, at least in a few runs I
	   did.  This has been fixed.

         - The patch adds some code to monitor how we're converging to a
	   solution, and to reset the iteration if necessary.  In the few
	   examples I've tried, this has made the convergence much more
	   robust.

         - The patch changes the start point of the iteration.  Before, we
	   started from a uniform gray ("x" was initialized to 0.0f
	   everywhere). Now, we start from the original image.  I found
	   this also helps the iteration be more robust, although
	   sometimes its quicker and sometimes its slower this way.

         - The clipping prevention code has been modified to be scale
	   independent.  A badly-scaled input file should no longer cause
	   a loss of dynamic range.

         Overall, the code runs sometimes a little more slowly and
	 sometimes about the same speed as before, although it is, I
	 believe, now far more robust and much more likely to converge.

	more at: http://groups.google.com/group/pfstools/browse_thread/thread/de2378af98ec6185/0dee5304fc14e99d?hl=en#0dee5304fc14e99d

	* mantiuk06: Ed Brambley's OpenMP patch:

	I've attached a second patch (which depends on the first one).
	This patch is a first attempt at parallelizing the code to run on
	OpenMP. I've tried to avoid assuming how many processors there
	will be, so it should scale to 4 or 8 processors quite happily.
	I'll let you know when I get my Quad core set up :-)

        There are probably more efficient ways to parallelize the code;
	for example, having only one or two "#pragma omp parallel"
	sections and lots of "#pragma omp for" or "#pragma omp sections"
	sections, but I've not tried to do this.  It happily uses 100% of
	both cores on the dual-core box I tried it on, and gives knocks
	about 20 seconds of a typical 1 minute tone mapping.
	
	* mantiuk06: fixed references to exp10f (missing in some compilers)

	* mantiuk06: added branch prediction optimizations (thanks to Lebed Dmytry)

	* mantiuk08: better handling of zero-values in HDR images - previously it could fail on some images

	* mantiuk08: removed some debug code

	* mantiuk08: added branch prediction optimizations
	
	* added: autoconf checks for OpenMP and some gcc optimization options
	
16/06/2008 PFSTMO 1.3

	* added: mantiuk08 - display adaptive tone mapping

	* mantiuk06: updated documentation, added --quiet option,
	        can now process more than one image in a loop
	
19/02/2008 PFSTMO 1.2

	* mantiuk06: fixed some serious bugs in the solver and subsampling
	        procedure the artifacts found in some odd-sized images
		should be gone

	* mantiuk06: faster up-/down-sampling functions make the
	        algorithm 25% faster. Thanks to Ed Brambley.

	* mantiuk06: performance improvements from Ed Brambley: lower
	        memory consumption, faster solver, better convergence,
	        additional scale parameter for the contrast equalization
	        method.
		
11/07/2007 PFSTMO 1.1

	* pfstmo-1.1: first release on sourceforge
		
	* small fixes with parameters (Miguel Granados)
		
	* reinhard05: renamed from reinhard04 to correct publication year.
		substantial corrections to the code
		  
	* fattal02: major bug fix,
		added noise level parameter (Przemyslaw Bazarnik),
	  manpage updates - explanation of parameters
		  
	* mantiuk06: new gradient domain operator
		implementation by Radoslaw Mantiuk
		
	* ashikhmin02: implementation temporarily dropped
		imposible to reproduce results from paper

05/12/2005

	* added temporal coherence to reinhard02
	
29/11/2005

	* fixed ansi compatibilities (fattal02)

14/12/2004

	* new algorithms
		ashikhmin02, fattal02

	* durand02
		corrected interpretation of compression factor
		percentile to estimate min and max in final mapping -> more stable

	* pattanaik02
		bug fix: local adaptation
		bug fix: r,g,b intensities normalization

	* reinhard02
		fix: memory leaks in building the pyramid

22/09/2004 PFStmo 1.0
	
	* first unofficial release

	* implemented algorithms
		drago03, durand02, pattanaik00, reinhard02, reinhard04