178 |
kaklik |
1 |
#ifndef GAUSS_H |
|
|
2 |
#define GAUSS_H |
|
|
3 |
|
|
|
4 |
#include <boost/multi_array.hpp> |
|
|
5 |
#include <cassert> |
|
|
6 |
#include <cmath> |
|
|
7 |
#include <deque> |
|
|
8 |
#include "image.h" |
|
|
9 |
#include "image_conv.h" |
|
|
10 |
#include "image_op.h" |
|
|
11 |
|
|
|
12 |
namespace mimas { |
|
|
13 |
|
|
|
14 |
/** @defgroup gauss Gaussian blur and Gauss gradient |
|
|
15 |
Methods for blurring images with a gauss-bell and gauss-gradient. |
|
|
16 |
The filter-parameter \f$\sigma\f$ can be choosen and the size of the filter |
|
|
17 |
is computed by choosing an upper bound for the approximation-error. |
|
|
18 |
|
|
|
19 |
The following example demonstrates how to blur an image: |
|
|
20 |
\include gauss_tool/main.cc |
|
|
21 |
|
|
|
22 |
@author Stuart Meikle (stu@stumeikle.org) |
|
|
23 |
@author Jan Wedekind (jan@wedesoft.de) |
|
|
24 |
@date Fri Apr 07 18:52:00 2006 |
|
|
25 |
@{ */ |
|
|
26 |
/** Compute gauss-bell. |
|
|
27 |
The values of the cells are computed by using differences of the integral |
|
|
28 |
of the gauss-function: |
|
|
29 |
\f$\displaystyle\int_{-r}^{+r}{\frac{1}{\sqrt{2\,\pi}\,\sigma}\,e^{-\displaystyle\frac{x^2}{2\,\sigma^2}}\,\mathrm{d}x}\ =\ \mathrm{erf}(\displaystyle\frac{r}{\sqrt{2}\,\sigma})\f$ |
|
|
30 |
|
|
|
31 |
The coefficients are normalised afterwards such that the sum of all |
|
|
32 |
elements of the filter is \f$1\f$. |
|
|
33 |
@param sigma Standard deviation of gauss-distribution. |
|
|
34 |
@param maxError Maximum error boundary |
|
|
35 |
(relative to range of pixel-values). |
|
|
36 |
@see gaussBlur */ |
|
|
37 |
template< typename T > |
|
|
38 |
std::deque< T > gaussBlurFilter( T sigma, T maxError = (T)( 1.0 / 256.0 ) ); |
|
|
39 |
|
|
|
40 |
/** Blur 2-D array. |
|
|
41 |
Perform gaussian blur on 2-D array. |
|
|
42 |
@param x Input array. |
|
|
43 |
@param sigma Standard deviation. |
|
|
44 |
@param maxError Maximum error boundary (relative to range of pixel-values). |
|
|
45 |
@see gaussBlurFilter */ |
|
|
46 |
template< typename T > |
|
|
47 |
boost::multi_array< T, 2 > gaussBlur |
|
|
48 |
( const boost::const_multi_array_ref< T, 2 > &x, T sigma, T maxError ); |
|
|
49 |
|
|
|
50 |
/** Blur image. |
|
|
51 |
Perform gaussian blur on 2-D image. |
|
|
52 |
@param x Input image. |
|
|
53 |
@param sigma Standard deviation. |
|
|
54 |
@param maxError Maximum error boundary (relative to range of pixel-values). |
|
|
55 |
@see gaussBlurFilter */ |
|
|
56 |
template< typename T > |
|
|
57 |
image< T > gaussBlur( const image< T > &x, T sigma, |
|
|
58 |
T maxError = (T)( 1.0 / 256.0 ) ) |
|
|
59 |
{ |
|
|
60 |
boost::const_multi_array_ref< T, 2 > data |
|
|
61 |
( x.rawData(), boost::extents[ x.getHeight() ][ x.getWidth() ] ); |
|
|
62 |
image< T > retVal; retVal.init( x.getWidth(), x.getHeight() ); |
|
|
63 |
boost::multi_array_ref< T, 2 > |
|
|
64 |
( retVal.rawData(), |
|
|
65 |
boost::extents[ retVal.getHeight() ][ retVal.getWidth() ] ) = |
|
|
66 |
gaussBlur< T >( data, sigma, maxError ); |
|
|
67 |
return retVal; |
|
|
68 |
} |
|
|
69 |
|
|
|
70 |
/** Compute gauss-gradient. |
|
|
71 |
The values of the cells are computed by using differences of the |
|
|
72 |
integral (the gauss-function): |
|
|
73 |
\f$\frac{1}{\sqrt{2\,\pi}\,\sigma}\,e^{-\displaystyle\frac{x^2}{2\,\sigma^2}}\,\mathrm{d}x\big\|_r^\infty\f$ |
|
|
74 |
|
|
|
75 |
The coefficients are normalised afterwards such that the sum of the square |
|
|
76 |
of all elements of the filter is \f$1\f$. |
|
|
77 |
@param sigma Standard deviation of gauss-distribution. |
|
|
78 |
@param maxError Maximum error boundary |
|
|
79 |
(relative to range of pixel-values). |
|
|
80 |
@see gaussBlur */ |
|
|
81 |
template< typename T > |
|
|
82 |
std::deque< T > gaussGradientFilter( T sigma, |
|
|
83 |
T maxError = (T)( 1.0 / 256.0 ) ); |
|
|
84 |
|
|
|
85 |
/** Take x-gradient of 2-D array. |
|
|
86 |
Compute gauss-gradient of 2-D array in x-direction. |
|
|
87 |
@param x Input array. |
|
|
88 |
@param sigma Standard deviation. |
|
|
89 |
@param maxError Maximum error boundary (relative to range of pixel-values). |
|
|
90 |
@see gaussGradientFilter */ |
|
|
91 |
template< typename T > |
|
|
92 |
boost::multi_array< T, 2 > gaussGradientX |
|
|
93 |
( const boost::const_multi_array_ref< T, 2 > &x, T sigma, T maxError ); |
|
|
94 |
|
|
|
95 |
/** Take x-gradient of 2-D image. |
|
|
96 |
Compute gauss-gradient of 2-D image in x-direction. |
|
|
97 |
@param x Input image. |
|
|
98 |
@param sigma Standard deviation. |
|
|
99 |
@param maxError Maximum error boundary (relative to range of pixel-values). |
|
|
100 |
@see gaussGradientFilter */ |
|
|
101 |
template< typename T > |
|
|
102 |
image< T > gaussGradientX( const image< T > &x, T sigma, |
|
|
103 |
T maxError = (T)( 1.0 / 256.0 ) ); |
|
|
104 |
|
|
|
105 |
/** Take y-gradient of 2-D array. |
|
|
106 |
Compute gauss-gradient of 2-D array in y-direction. |
|
|
107 |
@param x Input array. |
|
|
108 |
@param sigma Standard deviation. |
|
|
109 |
@param maxError Maximum error boundary (relative to range of pixel-values). |
|
|
110 |
@see gaussGradientFilter */ |
|
|
111 |
template< typename T > |
|
|
112 |
boost::multi_array< T, 2 > gaussGradientY |
|
|
113 |
( const boost::const_multi_array_ref< T, 2 > &x, T sigma, T maxError ); |
|
|
114 |
|
|
|
115 |
/** Take y-gradient of 2-D image. |
|
|
116 |
Compute gauss-gradient of 2-D image in y-direction. |
|
|
117 |
@param x Input image. |
|
|
118 |
@param sigma Standard deviation. |
|
|
119 |
@param maxError Maximum error boundary (relative to range of pixel-values). |
|
|
120 |
@see gaussGradientFilter */ |
|
|
121 |
template< typename T > |
|
|
122 |
image< T > gaussGradientY( const image< T > &x, T sigma, |
|
|
123 |
T maxError = (T)( 1.0 / 256.0 ) ); |
|
|
124 |
|
|
|
125 |
/** Square of gradient-norm. |
|
|
126 |
Compute square of gradient-norm for 2-D image |
|
|
127 |
@param im Input image. |
|
|
128 |
@param sigma Standard deviation. |
|
|
129 |
@param maxError Maximum error boundary (relative to range of pixel-values). |
|
|
130 |
@see gaussGradientFilter */ |
|
|
131 |
template< typename T > |
|
|
132 |
image< T > gaussGradientSqr( const image< T > &im, T sigma, |
|
|
133 |
T maxError = (T)( 1.0 / 256.0 ) ) |
|
|
134 |
{ |
|
|
135 |
image< T > |
|
|
136 |
gradX( gaussGradientX( im, sigma, maxError ) ), |
|
|
137 |
gradY( gaussGradientY( im, sigma, maxError ) ); |
|
|
138 |
return sumSquares( gradX, gradY ); |
|
|
139 |
} |
|
|
140 |
|
|
|
141 |
/** Gradient-norm. |
|
|
142 |
Compute gradient-norm for 2-D image |
|
|
143 |
@param im Input image. |
|
|
144 |
@param sigma Standard deviation. |
|
|
145 |
@param maxError Maximum error boundary (relative to range of pixel-values). |
|
|
146 |
@see gaussGradientFilter */ |
|
|
147 |
template< typename T > |
|
|
148 |
image< T > gaussGradientNorm( const image< T > &im, T sigma, |
|
|
149 |
T maxError = (T)( 1.0 / 256.0 ) ) |
|
|
150 |
{ |
|
|
151 |
return squareRoot( gaussGradientSqr( im, sigma, maxError ) ); |
|
|
152 |
} |
|
|
153 |
|
|
|
154 |
///@} |
|
|
155 |
|
|
|
156 |
}; |
|
|
157 |
|
|
|
158 |
#include "gauss.tcc" |
|
|
159 |
|
|
|
160 |
#endif |