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
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
use decoders::*;
use decoders::cfa::*;
use imageops;

/// All the data needed to process this raw image, including the image data itself as well
/// as all the needed metadata
#[derive(Debug, Clone)]
pub struct RawImage {
  /// camera make as encoded in the file
  pub make: String,
  /// camera model as encoded in the file
  pub model: String,
  /// make cleaned up to be consistent and short
  pub clean_make: String,
  /// model cleaned up to be consistent and short
  pub clean_model: String,
  /// width of the full image
  pub width: usize,
  /// height of the full image
  pub height: usize,
  /// number of components per pixel (1 for bayer, 3 for RGB images)
  pub cpp: usize,
  /// whitebalance coefficients encoded in the file in RGBE order
  pub wb_coeffs: [f32;4],
  /// image whitelevels in RGBE order
  pub whitelevels: [u16;4],
  /// image blacklevels in RGBE order
  pub blacklevels: [u16;4],
  /// matrix to convert XYZ to camera RGBE
  pub xyz_to_cam: [[f32;3];4],
  /// color filter array
  pub cfa: CFA,
  /// how much to crop the image to get all the usable area, order is top, right, bottom, left
  pub crops: [usize;4],
  /// image data itself, has width*height*cpp elements
  pub data: Vec<u16>,
}

/// A RawImage processed into a full RGB image with levels and gamma
///
/// The data is a Vec<f32> width width*height*3 elements, where each element is a value
/// between 0 and 1 with the intensity of the color channel
#[derive(Debug, Clone)]
pub struct RGBImage {
  pub width: usize,
  pub height: usize,
  pub data: Vec<f32>,
}

/// A RawImage processed into a full 8bit sRGB image with levels and gamma
///
/// The data is a Vec<u8> width width*height*3 elements, where each element is a value
/// between 0 and 255 with the intensity of the color channel with gamma applied
#[derive(Debug, Clone)]
pub struct SRGBImage {
  pub width: usize,
  pub height: usize,
  pub data: Vec<u8>,
}


impl RawImage {
  #[doc(hidden)] pub fn new(camera: &Camera, width: usize, height: usize, wb_coeffs: [f32;4], image: Vec<u16>) -> RawImage {
    let blacks = if camera.blackareah.1 != 0 || camera.blackareav.1 != 0 {
      let mut avg = [0 as f32; 4];
      let mut count = [0 as f32; 4];
      for row in camera.blackareah.0 .. camera.blackareah.0+camera.blackareah.1 {
        for col in 0..width {
          let color = camera.cfa.color_at(row,col);
          avg[color] += image[row*width+col] as f32;
          count[color] += 1.0;
        }
      }
      for row in 0..height {
        for col in camera.blackareav.0 .. camera.blackareav.0+camera.blackareav.1 {
          let color = camera.cfa.color_at(row,col);
          avg[color] += image[row*width+col] as f32;
          count[color] += 1.0;
        }
      }
      [(avg[0]/count[0]) as u16,
       (avg[1]/count[1]) as u16,
       (avg[2]/count[2]) as u16,
       (avg[3]/count[3]) as u16]
    } else {
      camera.blacklevels
    };

    RawImage {
      make: camera.make.clone(),
      model: camera.model.clone(),
      clean_make: camera.clean_make.clone(),
      clean_model: camera.clean_model.clone(),
      width: width,
      height: height,
      cpp: 1,
      wb_coeffs: wb_coeffs,
      data: image,
      blacklevels: blacks,
      whitelevels: camera.whitelevels,
      xyz_to_cam: camera.xyz_to_cam,
      cfa: camera.cfa.clone(),
      crops: camera.crops,
    }
  }

  /// Outputs the inverted matrix that converts pixels in the camera colorspace into
  /// XYZ components. Those can then be easily used to convert to Lab or a RGB output space
  pub fn cam_to_xyz(&self) -> [[f32;4];3] {
    let (cam_to_xyz, _) = self.xyz_matrix_and_neutralwb();
    cam_to_xyz
  }

  /// Not all cameras encode a whitebalance so in those cases just using a 6500K neutral one
  /// is a good compromise
  pub fn neutralwb(&self) -> [f32;4] {
    let (_, neutralwb) = self.xyz_matrix_and_neutralwb();
    [neutralwb[0]/neutralwb[1],
     neutralwb[1]/neutralwb[1],
     neutralwb[2]/neutralwb[1],
     neutralwb[3]/neutralwb[1]]
  }

  fn xyz_matrix_and_neutralwb(&self) -> ([[f32;4];3],[f32;4]) {
    let d65_white = [0.9547,1.0,1.08883];
    let rgb_to_xyz = [
    // sRGB D65
      [ 0.412453, 0.357580, 0.180423 ],
      [ 0.212671, 0.715160, 0.072169 ],
      [ 0.019334, 0.119193, 0.950227 ],
    ];

    // Multiply RGB matrix
    let mut rgb_to_cam = [[0.0;3];4];
    for i in 0..4 {
      for j in 0..3 {
        rgb_to_cam[i][j] = 0.0;
        for k in 0..3 {
          rgb_to_cam[i][j] += self.xyz_to_cam[i][k] * rgb_to_xyz[k][j];
        }
      }
    }

    let mut neutralwb = [0 as f32; 4];
    // Normalize rgb_to_cam so that rgb_to_cam * (1,1,1) is (1,1,1,1)
    for i in 0..4 {
      let mut num = 0.0;
      for j in 0..3 {
        num += rgb_to_cam[i][j];
      }
      for j in 0..3 {
        rgb_to_cam[i][j] = if num == 0.0 {
          0.0
        }  else {
          rgb_to_cam[i][j] / num
        };
      }
      neutralwb[i] = 1.0 / num;
    }

    let cam_to_rgb = self.pseudoinverse(rgb_to_cam);
    let mut cam_to_xyz = [[0.0;4];3];
    // Multiply RGB matrix and adjust white to get a cam_to_xyz
    for i in 0..3 {
      for j in 0..4 {
        cam_to_xyz[i][j] = 0.0;
        for k in 0..3 {
          cam_to_xyz[i][j] += cam_to_rgb[k][j] * rgb_to_xyz[i][k] / d65_white[i];
        }
      }
    }

    (cam_to_xyz, neutralwb)
  }

  fn pseudoinverse(&self, inm: [[f32;3];4]) -> [[f32;4];3] {
    let mut temp: [[f32;6];3] = [[0.0; 6];3];

    for i in 0..3 {
      for j in 0..6 {
        temp[i][j] = if j == i+3 { 1.0 } else { 0.0 };
      }
      for j in 0..3 {
        for k in 0..4 {
          temp[i][j] += inm[k][i] * inm[k][j];
        }
      }
    }

    for i in 0..3 {
      let mut num = temp[i][i];
      for j in 0..6 {
        temp[i][j] /= num;
      }
      for k in 0..3 {
        if k == i { continue }
        num = temp[k][i];
        for j in 0..6 {
          temp[k][j] -= temp[i][j] * num;
        }
      }
    }

    let mut out: [[f32;4];3] = [[0.0; 4];3];

    for i in 0..4 {
      for j in 0..3 {
        out[j][i] = 0.0;
        for k in 0..3 {
          out[j][i] += temp[j][k+3] * inm[i][k];
        }
      }
    }

    out
  }

  /// Convert the image to a sRGB image by doing a demosaic and applying levels,
  /// whitebalance, a base curve, color conversions and gamma.
  ///
  /// The maxwidth and maxheight values specify maximum dimensions for the final image. If
  /// the original image is smaller this will not scale up but otherwise you will get an
  /// image that is either maxwidth wide or maxheight tall and maintains the image ratio.
  /// Pass in maxwidth and maxheight as 0 if you want the maximum possible image size.
  pub fn to_rgb(&self, maxwidth: usize, maxheight: usize) -> Result<RGBImage,String> {
    let buffer = imageops::simple_decode(self, maxwidth, maxheight);

    Ok(RGBImage{
      width: buffer.width,
      height: buffer.height,
      data: buffer.data,
    })
  }

  /// Convert the image to an 8bit sRGB image by doing a demosaic and applying levels,
  /// whitebalance, a base curve, color conversions and gamma.
  ///
  /// The maxwidth and maxheight values specify maximum dimensions for the final image. If
  /// the original image is smaller this will not scale up but otherwise you will get an
  /// image that is either maxwidth wide or maxheight tall and maintains the image ratio.
  /// Pass in maxwidth and maxheight as 0 if you want the maximum possible image size.
  pub fn to_srgb(&self, maxwidth: usize, maxheight: usize) -> Result<SRGBImage,String> {
    let buffer = imageops::simple_decode(self, maxwidth, maxheight);
    let mut image = vec![0 as u8; buffer.width*buffer.height*3];
    for (o, i) in image.chunks_mut(1).zip(buffer.data.iter()) {
      o[0] = (i*255.0).max(0.0).min(255.0) as u8;
    }

    Ok(SRGBImage{
      width: buffer.width,
      height: buffer.height,
      data: image,
    })
  }

  /// Convert the image to a linear sRGB image by doing a demosaic and applying levels and
  /// whitebalance, a base curve and color conversions.
  ///
  /// The maxwidth and maxheight values specify maximum dimensions for the final image. If
  /// the original image is smaller this will not scale up but otherwise you will get an
  /// image that is either maxwidth wide or maxheight tall and maintains the image ratio.
  /// Pass in maxwidth and maxheight as 0 if you want the maximum possible image size.
  pub fn to_linear_rgb(&self, maxwidth: usize, maxheight: usize) -> Result<RGBImage,String> {
    let buffer = imageops::simple_decode_linear(self, maxwidth, maxheight);

    Ok(RGBImage{
      width: buffer.width,
      height: buffer.height,
      data: buffer.data,
    })
  }
}