imagepipe 0.5.0

An image processing pipeline
Documentation 
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

use crate::opbasics::*;

#[derive(Copy, Clone, Debug, Serialize, Deserialize)]
pub struct OpGoFloat {
  pub crop_top: usize,
  pub crop_right: usize,
  pub crop_bottom: usize,
  pub crop_left: usize,
  pub is_cfa: bool,
  pub blacklevels: [f32;4],
  pub whitelevels: [f32;4],
}

fn from_int4(arr: [u16;4]) -> [f32;4] {
  [arr[0] as f32, arr[1] as f32, arr[2] as f32, arr[3] as f32]
}

impl OpGoFloat {
  pub fn new(img: &ImageSource) -> OpGoFloat {
    match img {
      ImageSource::Raw(img) => {
        // Calculate the resulting width/height and top-left corner after crops
        OpGoFloat{
          crop_top:    img.crops[0],
          crop_right:  img.crops[1],
          crop_bottom: img.crops[2],
          crop_left:   img.crops[3],
          is_cfa: img.cfa.is_valid(),
          blacklevels: from_int4(img.blacklevels),
          whitelevels: from_int4(img.whitelevels),
        }
      },
      ImageSource::Other(_) => {
        OpGoFloat{
          crop_top:    0,
          crop_right:  0,
          crop_bottom: 0,
          crop_left:   0,
          is_cfa: false,
          blacklevels: [0.0, 0.0, 0.0, 0.0], // Unused
          whitelevels: [0.0, 0.0, 0.0, 0.0], // Unused
        }
      }
    }
  }
}

impl<'a> ImageOp<'a> for OpGoFloat {
  fn name(&self) -> &str {"gofloat"}
  fn run(&self, pipeline: &PipelineGlobals, _buf: Arc<OpBuffer>) -> Arc<OpBuffer> {
    // FIXME: Doing all the transforms with lookup tables instead of f32 calcs
    //        on every pixel is much faster

    match &pipeline.image {
      ImageSource::Raw(img) => {
        self.run_raw(img)
      },
      ImageSource::Other(img) => {
        self.run_other(img)
      }
    }
  }

  fn transform_forward(&mut self, width: usize, height: usize) -> (usize, usize) {
    let (_, _, width, height) = self.size_image(width, height);
    (width, height)
  }

  // We don't transform_reverse as image sizing is relative to the scaling done
  // at the demosaic step
}

impl OpGoFloat {
  fn size_image(&self, owidth: usize, oheight: usize) -> (usize, usize, usize, usize) {
    // Calculate x/y/width/height making sure we get at least a 10x10 "image" to not trip up
    // reasonable assumptions in later ops
    let x = cmp::min(self.crop_left, owidth-10);
    let y = cmp::min(self.crop_top, oheight-10);
    let width = owidth - cmp::min(self.crop_left + self.crop_right, owidth-10);
    let height = oheight - cmp::min(self.crop_top + self.crop_bottom, oheight-10);
    (x, y, width, height)
  }

  fn run_raw(&self, img: &RawImage) -> Arc<OpBuffer> {
    // Calculate the levels
    let mins = self.blacklevels;
    let ranges = self.whitelevels.iter().enumerate().map(|(i, &x)| {
      x - mins[i]
    }).collect::<Vec<f32>>();

    let owidth = img.width;
    let oheight = img.height;
    let (x, y, width, height) = self.size_image(owidth, oheight);

    Arc::new(match img.data {
      RawImageData::Integer(ref data) => {
        if img.cpp == 1 && !self.is_cfa {
          // We're in a monochrome image so turn it into RGB
          let mut out = OpBuffer::new(width, height, 4, true);
          out.mutate_lines(&(|line: &mut [f32], row| {
            for (o, i) in line.chunks_exact_mut(4).zip(data[owidth*(row+y)+x..].chunks_exact(1)) {
              let val = ((i[0] as f32 - mins[0]) / ranges[0]).min(1.0);
              o[0] = val;
              o[1] = val;
              o[2] = val;
              o[3] = 0.0;
            }
          }));
          out
        } else if img.cpp == 3 {
          // We're in an RGB image, turn it into four channel
          let mut out = OpBuffer::new(width, height, 4, false);
          out.mutate_lines(&(|line: &mut [f32], row| {
            for (o, i) in line.chunks_exact_mut(4).zip(data[(owidth*(row+y)+x)*3..].chunks_exact(3)) {
              o[0] = ((i[0] as f32 - mins[0]) / ranges[0]).min(1.0);
              o[1] = ((i[1] as f32 - mins[1]) / ranges[1]).min(1.0);
              o[2] = ((i[2] as f32 - mins[2]) / ranges[2]).min(1.0);
              o[3] = 0.0;
            }
          }));
          out
        } else {
          let mut out = OpBuffer::new(width, height, img.cpp, false);
          out.mutate_lines(&(|line: &mut [f32], row| {
            for (o, i) in line.chunks_exact_mut(1).zip(data[owidth*(row+y)+x..].chunks_exact(1)) {
              o[0] = ((i[0] as f32 - mins[0]) / ranges[0]).min(1.0);
            }
          }));
          out
        }
      },
      RawImageData::Float(ref data) => {
        if img.cpp == 1 && !self.is_cfa {
          // We're in a monochrome image so turn it into RGB
          let mut out = OpBuffer::new(width, height, 4, true);
          out.mutate_lines(&(|line: &mut [f32], row| {
            for (o, i) in line.chunks_exact_mut(4).zip(data[owidth*(row+y)+x..].chunks_exact(1)) {
              let val = ((i[0] as f32 - mins[0]) / ranges[0]).min(1.0);
              o[0] = val;
              o[1] = val;
              o[2] = val;
              o[3] = 0.0;
            }
          }));
          out
        } else if img.cpp == 3 {
          // We're in an RGB image, turn it into four channel
          let mut out = OpBuffer::new(width, height, 4, false);
          out.mutate_lines(&(|line: &mut [f32], row| {
            for (o, i) in line.chunks_exact_mut(4).zip(data[(owidth*(row+y)+x)*3..].chunks_exact(3)) {
              o[0] = ((i[0] as f32 - mins[0]) / ranges[0]).min(1.0);
              o[1] = ((i[1] as f32 - mins[1]) / ranges[1]).min(1.0);
              o[2] = ((i[2] as f32 - mins[2]) / ranges[2]).min(1.0);
              o[3] = 0.0;
            }
          }));
          out
        } else {
          let mut out = OpBuffer::new(width, height, img.cpp, false);
          out.mutate_lines(&(|line: &mut [f32], row| {
            for (o, i) in line.chunks_exact_mut(1).zip(data[owidth*(row+y)+x..].chunks_exact(1)) {
              o[0] = ((i[0] as f32 - mins[0]) / ranges[0]).min(1.0);
            }
          }));
          out
        }
      },
    })
  }

  fn run_other(&self, img: &OtherImage) -> Arc<OpBuffer> {
    let owidth = img.width() as usize;
    let oheight = img.height() as usize;
    let (x, y, width, height) = self.size_image(owidth, oheight);
    let mut out = OpBuffer::new(width, height, 4, false);
    let bits_per_channel = img.color().bits_per_pixel() / img.color().channel_count() as u16;

    if bits_per_channel == 8 {
      let data = img.to_rgb8().into_raw();
      out.mutate_lines(&(|line: &mut [f32], row| {
        for (o, i) in line.chunks_exact_mut(4).zip(data[(owidth*(row+y)+x)*3..].chunks_exact(3)) {
          o[0] = expand_srgb_gamma(input8bit(i[0]));
          o[1] = expand_srgb_gamma(input8bit(i[1]));
          o[2] = expand_srgb_gamma(input8bit(i[2]));
          o[3] = 0.0;
        }
      }));
    } else {
      let data = img.to_rgb16().into_raw();
      out.mutate_lines(&(|line: &mut [f32], row| {
        for (o, i) in line.chunks_exact_mut(4).zip(data[(owidth*(row+y)+x)*3..].chunks_exact(3)) {
          o[0] = input16bit(i[0]);
          o[1] = input16bit(i[1]);
          o[2] = input16bit(i[2]);
          o[3] = 0.0;
        }
      }));
    }

    Arc::new(out)
  }
}