demosaic 0.3.0

Demosaicing algorithms for Bayer and X-Trans CFA sensors
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
203
204
205
206
207
208
209
210
211

use crate::cfa::{CfaPattern, Channel};

/// Malvar-He-Cutler (2004) gradient-corrected linear demosaicing.
///
/// Uses 5×5 convolution kernels with integer coefficients to interpolate
/// missing color channels. Each CFA position has specific kernels that
/// incorporate gradient correction from the known channel, producing much
/// better results than simple bilinear averaging.
///
/// Reference: H.S. Malvar, L. He, R. Cutler, "High-Quality Linear
/// Interpolation for Demosaicing of Bayer-Patterned Color Images", ICASSP 2004.
pub fn demosaic(input: &[f32], width: usize, height: usize, cfa: &CfaPattern, output: &mut [f32]) {
    let npix = width * height;

    for y in 0..height {
        for x in 0..width {
            let idx = y * width + x;
            let ch = cfa.color_at(y, x);

            let r_off = 0;
            let g_off = npix;
            let b_off = 2 * npix;

            // Known channel: copy directly
            output[ch as usize * npix + idx] = input[idx];

            match ch {
                Channel::Red => {
                    output[g_off + idx] =
                        apply_kernel(input, width, height, y, x, &KERNEL_G_AT_RB, 8.0);
                    output[b_off + idx] =
                        apply_kernel(input, width, height, y, x, &KERNEL_RB_AT_OPPOSITE, 16.0);
                }
                Channel::Blue => {
                    output[g_off + idx] =
                        apply_kernel(input, width, height, y, x, &KERNEL_G_AT_RB, 8.0);
                    output[r_off + idx] =
                        apply_kernel(input, width, height, y, x, &KERNEL_RB_AT_OPPOSITE, 16.0);
                }
                Channel::Green => {
                    // Determine row type: is the adjacent pixel in this row R or B?
                    let adj_x = x ^ 1;
                    let adj_ch = cfa.color_at(y, adj_x);

                    if adj_ch == Channel::Red {
                        // Green in red row: R has same-row neighbors, B has same-column
                        output[r_off + idx] =
                            apply_kernel(input, width, height, y, x, &KERNEL_RB_AT_G_SAME_ROW, 16.0);
                        output[b_off + idx] =
                            apply_kernel(input, width, height, y, x, &KERNEL_RB_AT_G_SAME_COL, 16.0);
                    } else {
                        // Green in blue row: B has same-row neighbors, R has same-column
                        output[b_off + idx] =
                            apply_kernel(input, width, height, y, x, &KERNEL_RB_AT_G_SAME_ROW, 16.0);
                        output[r_off + idx] =
                            apply_kernel(input, width, height, y, x, &KERNEL_RB_AT_G_SAME_COL, 16.0);
                    }
                }
            }
        }
    }
}

#[inline]
fn apply_kernel(
    input: &[f32],
    width: usize,
    height: usize,
    y: usize,
    x: usize,
    kernel: &[i32; 25],
    divisor: f32,
) -> f32 {
    let mut sum = 0.0f32;
    let mut k = 0;
    for ky in -2i32..=2 {
        let ny = (y as i32 + ky).clamp(0, height as i32 - 1) as usize;
        for kx in -2i32..=2 {
            let nx = (x as i32 + kx).clamp(0, width as i32 - 1) as usize;
            sum += kernel[k] as f32 * input[ny * width + nx];
            k += 1;
        }
    }
    sum / divisor
}

// Green at Red or Blue location.
// Divisor: 8
#[rustfmt::skip]
const KERNEL_G_AT_RB: [i32; 25] = [
     0,  0, -1,  0,  0,
     0,  0,  2,  0,  0,
    -1,  2,  4,  2, -1,
     0,  0,  2,  0,  0,
     0,  0, -1,  0,  0,
];

// R/B at Green location where the same color has neighbors in the same ROW.
// e.g. Red at Green-in-Red-row, or Blue at Green-in-Blue-row.
// Divisor: 16
#[rustfmt::skip]
const KERNEL_RB_AT_G_SAME_ROW: [i32; 25] = [
     0,  0,  1,  0,  0,
     0, -2,  0, -2,  0,
    -2,  8, 10,  8, -2,
     0, -2,  0, -2,  0,
     0,  0,  1,  0,  0,
];

// R/B at Green location where the same color has neighbors in the same COLUMN.
// Transpose of KERNEL_RB_AT_G_SAME_ROW.
// Divisor: 16
#[rustfmt::skip]
const KERNEL_RB_AT_G_SAME_COL: [i32; 25] = [
     0,  0, -2,  0,  0,
     0, -2,  8, -2,  0,
     1,  0, 10,  0,  1,
     0, -2,  8, -2,  0,
     0,  0, -2,  0,  0,
];

// R at B location or B at R location (diagonal interpolation).
// Divisor: 16
#[rustfmt::skip]
const KERNEL_RB_AT_OPPOSITE: [i32; 25] = [
     0,  0, -3,  0,  0,
     0,  4,  0,  4,  0,
    -3,  0, 12,  0, -3,
     0,  4,  0,  4,  0,
     0,  0, -3,  0,  0,
];

#[cfg(test)]
mod tests {
    use alloc::vec;
    use super::*;

    #[test]
    fn known_channel_preserved() {
        for cfa in &[
            CfaPattern::bayer_rggb(),
            CfaPattern::bayer_bggr(),
            CfaPattern::bayer_grbg(),
            CfaPattern::bayer_gbrg(),
        ] {
            let w = 16;
            let h = 16;
            let mut input = vec![0.0f32; w * h];
            for y in 0..h {
                for x in 0..w {
                    input[y * w + x] = match cfa.color_at(y, x) {
                        Channel::Red => 0.8,
                        Channel::Green => 0.5,
                        Channel::Blue => 0.3,
                    };
                }
            }
            let mut output = vec![0.0f32; 3 * w * h];
            demosaic(&input, w, h, cfa, &mut output);

            for y in 0..h {
                for x in 0..w {
                    let idx = y * w + x;
                    let ch = cfa.color_at(y, x) as usize;
                    assert_eq!(
                        output[ch * w * h + idx], input[idx],
                        "known channel mismatch at ({y},{x}) ch={ch}"
                    );
                }
            }
        }
    }

    #[test]
    fn uniform_color_reconstructed() {
        for cfa in &[
            CfaPattern::bayer_rggb(),
            CfaPattern::bayer_bggr(),
            CfaPattern::bayer_grbg(),
            CfaPattern::bayer_gbrg(),
        ] {
            let w = 16;
            let h = 16;
            let mut input = vec![0.0f32; w * h];
            for y in 0..h {
                for x in 0..w {
                    input[y * w + x] = match cfa.color_at(y, x) {
                        Channel::Red => 0.8,
                        Channel::Green => 0.5,
                        Channel::Blue => 0.3,
                    };
                }
            }
            let mut output = vec![0.0f32; 3 * w * h];
            demosaic(&input, w, h, cfa, &mut output);

            // Interior pixels (skip 3-pixel border)
            for y in 3..h - 3 {
                for x in 3..w - 3 {
                    let idx = y * w + x;
                    let r = output[idx];
                    let g = output[w * h + idx];
                    let b = output[2 * w * h + idx];
                    assert!((r - 0.8).abs() < 0.05, "R at ({y},{x}) = {r}");
                    assert!((g - 0.5).abs() < 0.05, "G at ({y},{x}) = {g}");
                    assert!((b - 0.3).abs() < 0.05, "B at ({y},{x}) = {b}");
                }
            }
        }
    }
}