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
|
// SPDX-License-Identifier: LGPL-3.0-only
#ifndef RADLER_ALGORITHMS_MULTISCALE_MULTISCALE_TRANSFORMS_H_
#define RADLER_ALGORITHMS_MULTISCALE_MULTISCALE_TRANSFORMS_H_
#include <cmath>
#include <initializer_list>
#include <vector>
#include <aocommon/image.h>
#include <aocommon/uvector.h>
#include "settings.h"
namespace radler::algorithms::multiscale {
class MultiScaleTransforms {
public:
using Shape = radler::MultiscaleShape;
MultiScaleTransforms(size_t width, size_t height, Shape shape)
: _width(width), _height(height), _shape(shape) {}
MultiScaleTransforms(const MultiScaleTransforms&) = default;
MultiScaleTransforms(MultiScaleTransforms&&) = default;
MultiScaleTransforms& operator=(const MultiScaleTransforms&) = default;
MultiScaleTransforms& operator=(MultiScaleTransforms&&) = default;
void PrepareTransform(float* kernel, float scale);
void FinishTransform(float* image, const float* kernel);
void Transform(aocommon::Image& image, aocommon::Image& scratch,
float scale) {
std::vector<aocommon::Image> images(1, std::move(image));
Transform(images, scratch, scale);
image = std::move(images[0]);
}
void Transform(std::vector<aocommon::Image>& images, aocommon::Image& scratch,
float scale);
size_t Width() const { return _width; }
size_t Height() const { return _height; }
static float KernelIntegratedValue(float scaleInPixels, size_t maxN,
Shape shape) {
size_t n;
aocommon::Image kernel = MakeShapeFunction(scaleInPixels, n, maxN, shape);
float value = 0.0;
for (float& x : kernel) value += x;
return value;
}
static float KernelPeakValue(double scaleInPixels, size_t maxN, Shape shape) {
size_t n;
aocommon::Image kernel = MakeShapeFunction(scaleInPixels, n, maxN, shape);
return kernel[n / 2 + (n / 2) * n];
}
static void AddShapeComponent(float* image, size_t width, size_t height,
float scaleSizeInPixels, size_t x, size_t y,
float gain, Shape shape) {
size_t n;
aocommon::Image kernel =
MakeShapeFunction(scaleSizeInPixels, n, std::min(width, height), shape);
int left;
if (x > n / 2)
left = x - n / 2;
else
left = 0;
int top;
if (y > n / 2)
top = y - n / 2;
else
top = 0;
size_t right = std::min(x + (n + 1) / 2, width);
size_t bottom = std::min(y + (n + 1) / 2, height);
for (size_t yi = top; yi != bottom; ++yi) {
float* imagePtr = &image[yi * width];
const float* kernelPtr =
&kernel.Data()[(yi + n / 2 - y) * n + left + n / 2 - x];
for (size_t xi = left; xi != right; ++xi) {
imagePtr[xi] += *kernelPtr * gain;
++kernelPtr;
}
}
}
static aocommon::Image MakeShapeFunction(float scaleSizeInPixels, size_t& n,
size_t maxN, Shape shape) {
switch (shape) {
default:
case Shape::kTaperedQuadraticShape:
return makeTaperedQuadraticShapeFunction(scaleSizeInPixels, n);
case Shape::kGaussianShape:
return makeGaussianFunction(scaleSizeInPixels, n, maxN);
}
}
aocommon::Image MakeShapeFunction(float scaleSizeInPixels, size_t& n) {
return MakeShapeFunction(scaleSizeInPixels, n, std::min(_width, _height),
_shape);
}
static float GaussianSigma(float scaleSizeInPixels) {
return scaleSizeInPixels * (3.0 / 16.0);
}
private:
size_t _width, _height;
Shape _shape;
static size_t taperedQuadraticKernelSize(double scaleInPixels) {
return size_t(ceil(scaleInPixels * 0.5) * 2.0) + 1;
}
static aocommon::Image makeTaperedQuadraticShapeFunction(
double scaleSizeInPixels, size_t& n) {
n = taperedQuadraticKernelSize(scaleSizeInPixels);
aocommon::Image output(n, n);
taperedQuadraticShapeFunction(n, output, scaleSizeInPixels);
return output;
}
static aocommon::Image makeGaussianFunction(double scaleSizeInPixels,
size_t& n, size_t maxN) {
float sigma = GaussianSigma(scaleSizeInPixels);
n = int(ceil(sigma * 12.0 / 2.0)) * 2 + 1; // bounding box of 12 sigma
if (n > maxN) {
n = maxN;
if ((n % 2) == 0 && n > 0) --n;
}
if (n < 1) n = 1;
if (sigma == 0.0) {
sigma = 1.0;
n = 1;
}
aocommon::Image output(n, n);
const float mu = int(n / 2);
const float twoSigmaSquared = 2.0 * sigma * sigma;
float sum = 0.0;
float* outputPtr = output.Data();
aocommon::UVector<float> gaus(n);
for (int i = 0; i != int(n); ++i) {
float vI = float(i) - mu;
gaus[i] = std::exp(-vI * vI / twoSigmaSquared);
}
for (int y = 0; y != int(n); ++y) {
for (int x = 0; x != int(n); ++x) {
*outputPtr = gaus[x] * gaus[y];
sum += *outputPtr;
++outputPtr;
}
}
float normFactor = 1.0 / sum;
output *= normFactor;
return output;
}
static void taperedQuadraticShapeFunction(size_t n, aocommon::Image& output2d,
double scaleSizeInPixels) {
if (scaleSizeInPixels == 0.0)
output2d[0] = 1.0;
else {
float sum = 0.0;
float* outputPtr = output2d.Data();
for (int y = 0; y != int(n); ++y) {
float dy = y - 0.5 * (n - 1);
float dydy = dy * dy;
for (int x = 0; x != int(n); ++x) {
float dx = x - 0.5 * (n - 1);
float r = std::sqrt(dx * dx + dydy);
*outputPtr =
hannWindowFunction(r, n) * shapeFunction(r / scaleSizeInPixels);
sum += *outputPtr;
++outputPtr;
}
}
float normFactor = 1.0 / sum;
output2d *= normFactor;
}
}
static float hannWindowFunction(float x, size_t n) {
return (x * 2 <= n + 1)
? (0.5 * (1.0 + std::cos(2.0 * M_PI * x / double(n + 1))))
: 0.0;
}
static float shapeFunction(float x) {
return (x < 1.0) ? (1.0 - x * x) : 0.0;
}
};
} // namespace radler::algorithms::multiscale
#endif // RADLER_ALGORITHMS_MULTISCALE_MULTISCALE_TRANSFORMS_H_
|
