Tesseract  3.02
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Groups Pages
feature_chebyshev.cpp
Go to the documentation of this file.
1 /**********************************************************************
2  * File: feature_chebyshev.cpp
3  * Description: Implementation of the Chebyshev coefficients Feature Class
4  * Author: Ahmad Abdulkader
5  * Created: 2008
6  *
7  * (C) Copyright 2008, Google Inc.
8  ** Licensed under the Apache License, Version 2.0 (the "License");
9  ** you may not use this file except in compliance with the License.
10  ** You may obtain a copy of the License at
11  ** http://www.apache.org/licenses/LICENSE-2.0
12  ** Unless required by applicable law or agreed to in writing, software
13  ** distributed under the License is distributed on an "AS IS" BASIS,
14  ** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
15  ** See the License for the specific language governing permissions and
16  ** limitations under the License.
17  *
18  **********************************************************************/
19 
20 #include <stdio.h>
21 #include <stdlib.h>
22 #include <math.h>
23 #include <string>
24 #include <vector>
25 #include <algorithm>
26 #include "feature_base.h"
27 #include "feature_chebyshev.h"
28 #include "cube_utils.h"
29 #include "const.h"
30 #include "char_samp.h"
31 
32 #ifdef _WIN32
33 #ifndef M_PI
34 #define M_PI 3.14159265358979323846
35 #endif
36 #endif
37 
38 namespace tesseract {
39 
41  : FeatureBase(params) {
42 }
43 
45 }
46 
47 // Render a visualization of the features to a CharSamp.
48 // This is mainly used by visual-debuggers
50  return char_samp;
51 }
52 
53 // Compute Chebyshev coefficients for the specified vector
54 void FeatureChebyshev::ChebyshevCoefficients(const vector<float> &input,
55  int coeff_cnt, float *coeff) {
56  // re-sample function
57  int input_range = (input.size() - 1);
58  vector<float> resamp(coeff_cnt);
59  for (int samp_idx = 0; samp_idx < coeff_cnt; samp_idx++) {
60  // compute sampling position
61  float samp_pos = input_range *
62  (1 + cos(M_PI * (samp_idx + 0.5) / coeff_cnt)) / 2;
63  // interpolate
64  int samp_start = static_cast<int>(samp_pos);
65  int samp_end = static_cast<int>(samp_pos + 0.5);
66  float func_delta = input[samp_end] - input[samp_start];
67  resamp[samp_idx] = input[samp_start] +
68  ((samp_pos - samp_start) * func_delta);
69  }
70  // compute the coefficients
71  float normalizer = 2.0 / coeff_cnt;
72  for (int coeff_idx = 0; coeff_idx < coeff_cnt; coeff_idx++, coeff++) {
73  double sum = 0.0;
74  for (int samp_idx = 0; samp_idx < coeff_cnt; samp_idx++) {
75  sum += resamp[samp_idx] * cos(M_PI * coeff_idx * (samp_idx + 0.5) /
76  coeff_cnt);
77  }
78  (*coeff) = (normalizer * sum);
79  }
80 }
81 
82 // Compute the features of a given CharSamp
83 bool FeatureChebyshev::ComputeFeatures(CharSamp *char_samp, float *features) {
84  return ComputeChebyshevCoefficients(char_samp, features);
85 }
86 
87 // Compute the Chebyshev coefficients of a given CharSamp
89  float *features) {
90  if (char_samp->NormBottom() <= 0) {
91  return false;
92  }
93  unsigned char *raw_data = char_samp->RawData();
94  int stride = char_samp->Stride();
95  // compute the height of the word
96  int word_hgt = (255 * (char_samp->Top() + char_samp->Height()) /
97  char_samp->NormBottom());
98  // compute left & right profiles
99  vector<float> left_profile(word_hgt, 0.0);
100  vector<float> right_profile(word_hgt, 0.0);
101  unsigned char *line_data = raw_data;
102  for (int y = 0; y < char_samp->Height(); y++, line_data += stride) {
103  int min_x = char_samp->Width();
104  int max_x = -1;
105  for (int x = 0; x < char_samp->Width(); x++) {
106  if (line_data[x] == 0) {
107  UpdateRange(x, &min_x, &max_x);
108  }
109  }
110  left_profile[char_samp->Top() + y] =
111  1.0 * (min_x == char_samp->Width() ? 0 : (min_x + 1)) /
112  char_samp->Width();
113  right_profile[char_samp->Top() + y] =
114  1.0 * (max_x == -1 ? 0 : char_samp->Width() - max_x) /
115  char_samp->Width();
116  }
117 
118  // compute top and bottom profiles
119  vector<float> top_profile(char_samp->Width(), 0);
120  vector<float> bottom_profile(char_samp->Width(), 0);
121  for (int x = 0; x < char_samp->Width(); x++) {
122  int min_y = word_hgt;
123  int max_y = -1;
124  line_data = raw_data;
125  for (int y = 0; y < char_samp->Height(); y++, line_data += stride) {
126  if (line_data[x] == 0) {
127  UpdateRange(y + char_samp->Top(), &min_y, &max_y);
128  }
129  }
130  top_profile[x] = 1.0 * (min_y == word_hgt ? 0 : (min_y + 1)) / word_hgt;
131  bottom_profile[x] = 1.0 * (max_y == -1 ? 0 : (word_hgt - max_y)) / word_hgt;
132  }
133 
134  // compute the chebyshev coefficients of each profile
135  ChebyshevCoefficients(left_profile, kChebychevCoefficientCnt, features);
137  features + kChebychevCoefficientCnt);
139  features + (2 * kChebychevCoefficientCnt));
141  features + (3 * kChebychevCoefficientCnt));
142  return true;
143 }
144 } // namespace tesseract