Debian Codegraph @

upstream/debian-unstable/man1/pnmnlfilt.1

raw
1 \
2 .\" This man page was generated by the Netpbm tool 'makeman' from HTML source.
3 .\" Do not hand-hack it! If you have bug fixes or improvements, please find
4 .\" the corresponding HTML page on the Netpbm website, generate a patch
5 .\" against that, and send it to the Netpbm maintainer.
6 .TH "Pnmnlfilt User Manual" 1 "24 October 2006" "netpbm documentation"
7
8 .SH NAME
9
10 pnmnlfilt - non-linear filters: smooth, alpha trim mean, optimal
11 estimation smoothing, edge enhancement.
12
13 .UN synopsis
14 .SH SYNOPSIS
15
16 \fBpnmnlfilt\fP
17 \fIalpha\fP
18 \fIradius\fP
19 [\fIpnmfile\fP]
20
21 .UN description
22 .SH DESCRIPTION
23 .PP
24 This program is part of
25 .BR "Netpbm" (1)\c
26 \&.
27 .PP
28 \fBpnmnlfilt\fP produces an output image where the pixels are a
29 summary of multiple pixels near the corresponding location in an input
30 image.
31 .PP
32 This program works on multi-image streams.
33 .PP
34 This is something of a swiss army knife filter. It has 3 distinct
35 operating modes. In all of the modes \fBpnmnlfilt\fP examines each
36 pixel in the image and processes it according to the values of it and
37 its surrounding pixels. Rather than using a square block of
38 surrounding pixels (e.g. the subject pixel and its 8 immediate
39 neighbors, in a 3x3 square), \fBpnmnlfilt\fP uses 7 hexagonal areas.
40 You choose the size of the hexagons with the radius parameter. A
41 radius value of 1/3 means that the 7 hexagons essentially fit into the
42 subject pixel (ie. there will be no filtering effect). A radius
43 value of 1.0 means that the 7 hexagons essentially cover the 3x3
44 immediate neighbor square.
45 .PP
46 Your choice of "alpha" parameter selects among the three
47 modes.
48
49 .UN alphatrimmedmean
50 .SS
51 Alpha trimmed mean filter (0.0 <= alpha <= 0.5)
52 .PP
53 The value of the center pixel will be replaced by the mean of
54 the 7 hexagon values, but the 7 values are sorted by size and the top
55 and bottom alpha portion of the 7 are excluded from the mean. This
56 implies that an alpha value of 0.0 gives the same sort of output as a
57 normal convolution (ie. averaging or smoothing filter), where radius
58 will determine the "strength" of the filter. A good value to
59 start from for subtle filtering is alpha = 0.0, radius = 0.55 For a
60 more blatant effect, try alpha 0.0 and radius 1.0
61 .PP
62 An alpha value of 0.5 will cause the median value of the 7 hexagons
63 to be used to replace the center pixel value. This sort of filter is
64 good for eliminating "pop" or single pixel noise from an
65 image without spreading the noise out or smudging features on the
66 image. Judicious use of the radius parameter will fine tune the
67 filtering. Intermediate values of alpha give effects somewhere between
68 smoothing and "pop" noise reduction. For subtle filtering
69 try starting with values of alpha = 0.4, radius = 0.6 For a more
70 blatant effect try alpha = 0.5, radius = 1.0
71
72 .UN optimalestsmooth
73 .SS
74 Optimal estimation smoothing. (1.0 <= alpha <= 2.0)
75 .PP
76 This type of filter applies a smoothing filter adaptively over the
77 image. For each pixel the variance of the surrounding hexagon values
78 is calculated, and the amount of smoothing is made inversely
79 proportional to it. The idea is that if the variance is small then it
80 is due to noise in the image, while if the variance is large, it is
81 because of "wanted" image features. As usual the radius
82 parameter controls the effective radius, but it probably advisable to
83 leave the radius between 0.8 and 1.0 for the variance calculation to
84 be meaningful. The alpha parameter sets the noise threshold, over
85 which less smoothing will be done. This means that small values of
86 alpha will give the most subtle filtering effect, while large values
87 will tend to smooth all parts of the image. You could start with
88 values like alpha = 1.2, radius = 1.0 and try increasing or decreasing
89 the alpha parameter to get the desired effect. This type of filter is
90 best for filtering out dithering noise in both bitmap and color
91 images.
92
93 .UN edgeenhance
94 .SS Edge enhancement. (-0.1 >= alpha >= -0.9)
95 .PP
96 This is the opposite type of filter to the smoothing filter. It
97 enhances edges. The alpha parameter controls the amount of edge
98 enhancement, from subtle (-0.1) to blatant (-0.9). The radius
99 parameter controls the effective radius as usual, but useful values
100 are between 0.5 and 0.9. Try starting with values of alpha = 0.3,
101 radius = 0.8
102
103 .UN combination
104 .SS Combination use.
105 .PP
106 The various modes of \fBpnmnlfilt\fP can be used one after the
107 other to get the desired result. For instance to turn a monochrome
108 dithered image into a grayscale image you could try one or two passes
109 of the smoothing filter, followed by a pass of the optimal estimation
110 filter, then some subtle edge enhancement. Note that using edge
111 enhancement is only likely to be useful after one of the non-linear
112 filters (alpha trimmed mean or optimal estimation filter), as edge
113 enhancement is the direct opposite of smoothing.
114 .PP
115 For reducing color quantization noise in images (ie. turning .gif
116 files back into 24 bit files) you could try a pass of the optimal
117 estimation filter (alpha 1.2, radius 1.0), a pass of the median filter
118 (alpha 0.5, radius 0.55), and possibly a pass of the edge enhancement
119 filter. Several passes of the optimal estimation filter with
120 declining alpha values are more effective than a single pass with a
121 large alpha value. As usual, there is a tradeoff between filtering
122 effectiveness and losing detail. Experimentation is encouraged.
123
124 .UN options
125 .SH OPTIONS
126 .PP
127 There are no command line options defined specifically
128 for \fBpnmnlfilt\fP, but it recognizes the options common to all
129 programs based on libnetpbm (See
130 .UR index.html#commonoptions
131 Common Options
132 .UE
133 \&.)
134
135 .UN references
136 .SH REFERENCES
137 .PP
138 The alpha-trimmed mean filter is based on the description in IEEE
139 CG&A May 1990 Page 23 by Mark E. Lee and Richard A. Redner, and
140 has been enhanced to allow continuous alpha adjustment.
141 .PP
142 The optimal estimation filter is taken from an article
143 "Converting Dithered Images Back to Gray Scale" by Allen
144 Stenger, Dr Dobb's Journal, November 1992, and this article references
145 "Digital Image Enhancement and Noise Filtering by Use of Local
146 Statistics", Jong-Sen Lee, IEEE Transactions on Pattern Analysis
147 and Machine Intelligence, March 1980.
148 .PP
149 The edge enhancement details are from
150 .BR "pgmenhance" (1)\c
151 \&, which is taken from Philip
152 R. Thompson's "xim" program, which in turn took it from
153 section 6 of "Digital Halftones by Dot Diffusion",
154 D. E. Knuth, ACM Transaction on Graphics Vol. 6, No. 4, October 1987,
155 which in turn got it from two 1976 papers by J. F. Jarvis et. al.
156
157 .UN parameters
158 .SH PARAMETERS
159 .PP
160 The parameters are:
161
162
163 .TP
164 \fIalpha\fP
165 The alpha value (described above), in decimal. May be fractional.
166
167 .TP
168 \fIradius\fP
169 The radius (described above), in decimal. May be fractional.
170
171
172 .UN seealso
173 .SH SEE ALSO
174 .BR "pgmenhance" (1)\c
175 \&,
176 .BR "pnmconvol" (1)\c
177 \&,
178 .BR "pnm" (1)\c
179 \&
180
181 .UN author
182 .SH AUTHOR
183
184 Graeme W. Gill \[email protected]\fP
185 .SH DOCUMENT SOURCE
186 This manual page was generated by the Netpbm tool 'makeman' from HTML
187 source. The master documentation is at
188 .IP
189 .B http://netpbm.sourceforge.net/doc/pnmnlfilt.html
190 .PP