1Pnmnlfilt User Manual(0) Pnmnlfilt User Manual(0)
2
6 pnmnlfilt - non-linear filters: smooth, alpha trim mean, optimal esti‐
7 mation smoothing, edge enhancement.
8
11 pnmnlfilt alpha radius [pnmfile]
12
15 This program is part of Netpbm(1).
16 pnmnlfilt produces an output image where the pixels are a summary of
18 multiple pixels near the corresponding location in an input image.
19 This program works on multi-image streams.
21 This is something of a swiss army knife filter. It has 3 distinct
23 operating modes. In all of the modes pnmnlfilt examines each pixel in
24 the image and processes it according to the values of it and its sur‐
25 rounding pixels. Rather than using a square block of surrounding pix‐
26 els (e.g. the subject pixel and its 8 immediate neighbors, in a 3x3
27 square), pnmnlfilt uses 7 hexagonal areas. You choose the size of the
28 hexagons with the radius parameter. A radius value of 1/3 means that
29 the 7 hexagons essentially fit into the subject pixel (ie. there will
30 be no filtering effect). A radius value of 1.0 means that the 7
31 hexagons essentially cover the 3x3 immediate neighbor square.
32 Your choice of 'alpha' parameter selects among the three modes.
34 Alpha trimmed mean filter (0.0 <= alpha <= 0.5)
37 The value of the center pixel will be replaced by the mean of the 7
38 hexagon values, but the 7 values are sorted by size and the top and
39 bottom alpha portion of the 7 are excluded from the mean. This implies
40 that an alpha value of 0.0 gives the same sort of output as a normal
41 convolution (ie. averaging or smoothing filter), where radius will
42 determine the 'strength' of the filter. A good value to start from for
43 subtle filtering is alpha = 0.0, radius = 0.55 For a more blatant
44 effect, try alpha 0.0 and radius 1.0
45 An alpha value of 0.5 will cause the median value of the 7 hexagons to
47 be used to replace the center pixel value. This sort of filter is good
48 for eliminating 'pop' or single pixel noise from an image without
49 spreading the noise out or smudging features on the image. Judicious
50 use of the radius parameter will fine tune the filtering. Intermediate
51 values of alpha give effects somewhere between smoothing and 'pop'
52 noise reduction. For subtle filtering try starting with values of alpha
53 = 0.4, radius = 0.6 For a more blatant effect try alpha = 0.5, radius =
54 1.0
55 Optimal estimation smoothing. (1.0 <= alpha <= 2.0)
58 This type of filter applies a smoothing filter adaptively over the
59 image. For each pixel the variance of the surrounding hexagon values
60 is calculated, and the amount of smoothing is made inversely propor‐
61 tional to it. The idea is that if the variance is small then it is due
62 to noise in the image, while if the variance is large, it is because of
63 'wanted' image features. As usual the radius parameter controls the
64 effective radius, but it probably advisable to leave the radius between
65 0.8 and 1.0 for the variance calculation to be meaningful. The alpha
66 parameter sets the noise threshold, over which less smoothing will be
67 done. This means that small values of alpha will give the most subtle
68 filtering effect, while large values will tend to smooth all parts of
69 the image. You could start with values like alpha = 1.2, radius = 1.0
70 and try increasing or decreasing the alpha parameter to get the desired
71 effect. This type of filter is best for filtering out dithering noise
72 in both bitmap and color images.
73 Edge enhancement. (-0.1 >= alpha >= -0.9)
76 This is the opposite type of filter to the smoothing filter. It
77 enhances edges. The alpha parameter controls the amount of edge
78 enhancement, from subtle (-0.1) to blatant (-0.9). The radius parameter
79 controls the effective radius as usual, but useful values are between
80 0.5 and 0.9. Try starting with values of alpha = 0.3, radius = 0.8
81 Combination use.
84 The various modes of pnmnlfilt can be used one after the other to get
85 the desired result. For instance to turn a monochrome dithered image
86 into a grayscale image you could try one or two passes of the smoothing
87 filter, followed by a pass of the optimal estimation filter, then some
88 subtle edge enhancement. Note that using edge enhancement is only
89 likely to be useful after one of the non-linear filters (alpha trimmed
90 mean or optimal estimation filter), as edge enhancement is the direct
91 opposite of smoothing.
92 For reducing color quantization noise in images (ie. turning .gif files
94 back into 24 bit files) you could try a pass of the optimal estimation
95 filter (alpha 1.2, radius 1.0), a pass of the median filter (alpha 0.5,
96 radius 0.55), and possibly a pass of the edge enhancement filter. Sev‐
97 eral passes of the optimal estimation filter with declining alpha val‐
98 ues are more effective than a single pass with a large alpha value. As
99 usual, there is a tradeoff between filtering effectiveness and loosing
100 detail. Experimentation is encouraged.
101
104 The alpha-trimmed mean filter is based on the description in IEEE CG&A
105 May 1990 Page 23 by Mark E. Lee and Richard A. Redner, and has been
106 enhanced to allow continuous alpha adjustment.
107 The optimal estimation filter is taken from an article 'Converting
109 Dithered Images Back to Gray Scale' by Allen Stenger, Dr Dobb's Jour‐
110 nal, November 1992, and this article references 'Digital Image Enhance‐
111 ment and Noise Filtering by Use of Local Statistics', Jong-Sen Lee,
112 IEEE Transactions on Pattern Analysis and Machine Intelligence, March
113 1980.
114 The edge enhancement details are from pgmenhance(1),whichistaken‐
116 fromPhilip R. Thompson's 'xim' program, which in turn took it from sec‐
117 tion 6 of 'Digital Halftones by Dot Diffusion', D. E. Knuth, ACM Trans‐
118 action on Graphics Vol. 6, No. 4, October 1987, which in turn got it
119 from two 1976 papers by J. F. Jarvis et. al.
120 The parameters are:
125 alpha The alpha value (described above), in decimal. May be frac‐
126 tional.
127 radius The radius (described above), in decimal. May be fractional.
130
134 pgmenhance(1), pnmconvol(1), pnm(1)
135
138 Graeme W. Gill graeme@labtam.oz.au
139netpbm documentation 24 October 2006 Pnmnlfilt User Manual(0)