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
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
use crate::attr::{Attributes, ControlFlow};
use crate::error::*;
use crate::hist::{FixedColorsSet, HistogramInternal};
use crate::image::Image;
use crate::kmeans::Kmeans;
use crate::mediancut::mediancut;
use crate::pal::{PalIndex, PalF, PalLen, PalPop, Palette, LIQ_WEIGHT_MSE, MAX_COLORS, MAX_TRANSP_A, RGBA};
use crate::remap::{mse_to_standard_mse, DitherMapMode, Remapped};
use crate::seacow::RowBitmapMut;
use crate::OrdFloat;
use arrayvec::ArrayVec;
use std::cmp::Reverse;
use std::fmt;
use std::mem::MaybeUninit;
pub struct QuantizationResult {
remapped: Option<Box<Remapped>>,
pub(crate) palette: PalF,
progress_callback: Option<Box<dyn Fn(f32) -> ControlFlow + Send + Sync>>,
pub(crate) int_palette: Palette,
pub(crate) dither_level: f32,
pub(crate) gamma: f64,
pub(crate) palette_error: Option<f64>,
pub(crate) min_posterization_output: u8,
pub(crate) use_dither_map: DitherMapMode,
pub(crate) single_threaded_dithering: bool,
}
impl QuantizationResult {
pub(crate) fn new(attr: &Attributes, hist: HistogramInternal, freeze_result_colors: bool, fixed_colors: &FixedColorsSet, gamma: f64) -> Result<Self, Error> {
if attr.progress(attr.progress_stage1 as f32) { return Err(Aborted); }
let (max_mse, target_mse, target_mse_is_zero) = attr.target_mse(hist.items.len());
let (mut palette, palette_error) = find_best_palette(attr, target_mse, target_mse_is_zero, max_mse, hist, fixed_colors)?;
if freeze_result_colors {
palette.iter_mut().for_each(|(_, p)| *p = p.to_fixed());
}
if attr.progress(attr.progress_stage1 as f32 + attr.progress_stage2 as f32 + attr.progress_stage3 as f32 * 0.95) {
return Err(Aborted);
}
if let (Some(palette_error), Some(max_mse)) = (palette_error, max_mse) {
if palette_error > max_mse {
attr.verbose_print(format!(
" image degradation MSE={:0.3} (Q={}) exceeded limit of {:0.3} ({})",
mse_to_standard_mse(palette_error),
mse_to_quality(palette_error),
mse_to_standard_mse(max_mse),
mse_to_quality(max_mse)
));
return Err(QualityTooLow);
}
}
sort_palette(attr, &mut palette);
Ok(Self {
palette,
gamma,
palette_error,
min_posterization_output: attr.min_posterization(),
use_dither_map: attr.use_dither_map,
remapped: None,
progress_callback: None,
int_palette: Palette {
count: 0,
entries: [Default::default(); MAX_COLORS],
},
dither_level: 1.,
single_threaded_dithering: attr.single_threaded_dithering,
})
}
pub(crate) fn write_remapped_image_rows_internal(&mut self, image: &mut Image, output_pixels: RowBitmapMut<'_, MaybeUninit<PalIndex>>) -> Result<(), Error> {
if image.edges.is_none() && image.dither_map.is_none() && self.use_dither_map != DitherMapMode::None {
image.contrast_maps()?;
}
self.remapped = Some(Box::new(Remapped::new(self, image, output_pixels)?));
Ok(())
}
pub fn set_dithering_level(&mut self, value: f32) -> Result<(), Error> {
if !(0. ..=1.).contains(&value) {
return Err(ValueOutOfRange);
}
self.remapped = None;
self.dither_level = value;
Ok(())
}
pub fn set_output_gamma(&mut self, value: f64) -> Result<(), Error> {
if value <= 0. || value >= 1. {
return Err(ValueOutOfRange);
}
self.remapped = None;
self.gamma = value;
Ok(())
}
#[inline]
#[must_use]
pub fn output_gamma(&self) -> f64 {
self.gamma
}
#[must_use]
pub fn quantization_quality(&self) -> Option<u8> {
self.palette_error.map(mse_to_quality)
}
#[must_use]
pub fn quantization_error(&self) -> Option<f64> {
self.palette_error.map(mse_to_standard_mse)
}
pub fn remapping_error(&self) -> Option<f64> {
self.remapped.as_ref()
.and_then(|re| re.palette_error)
.or(self.palette_error)
.map(mse_to_standard_mse)
}
pub fn remapping_quality(&self) -> Option<u8> {
self.remapped.as_ref()
.and_then(|re| re.palette_error)
.or(self.palette_error)
.map(mse_to_quality)
}
#[inline]
pub fn palette(&mut self) -> &[RGBA] {
self.int_palette().as_slice()
}
pub(crate) fn int_palette(&mut self) -> &Palette {
match self.remapped.as_ref() {
Some(remap) => {
debug_assert!(remap.int_palette.count > 0);
&remap.int_palette
}
None => {
if self.int_palette.count == 0 {
self.int_palette = self.palette.make_int_palette(self.gamma, self.min_posterization_output);
}
&self.int_palette
},
}
}
#[inline(always)]
pub fn set_progress_callback<F: Fn(f32) -> ControlFlow + Sync + Send + 'static>(&mut self, callback: F) {
self.progress_callback = Some(Box::new(callback));
}
pub(crate) fn remap_progress(&self, percent: f32) -> bool {
if let Some(cb) = &self.progress_callback {
cb(percent) == ControlFlow::Break
} else {
false
}
}
pub fn remapped(&mut self, image: &mut Image<'_>) -> Result<(Vec<RGBA>, Vec<PalIndex>), Error> {
let len = image.width() * image.height();
unsafe {
let mut buf = Vec::new();
buf.try_reserve_exact(len)?;
let uninit_slice = &mut buf.spare_capacity_mut()[..len];
self.remap_into(image, uninit_slice)?;
buf.set_len(len);
Ok((self.palette_vec(), buf))
}
}
#[inline]
pub fn remap_into(&mut self, image: &mut Image<'_>, output_buf: &mut [MaybeUninit<PalIndex>]) -> Result<(), Error> {
let required_size = (image.width()) * (image.height());
let output_buf = output_buf.get_mut(0..required_size).ok_or(BufferTooSmall)?;
let rows = RowBitmapMut::new_contiguous(output_buf, image.width());
self.write_remapped_image_rows_internal(image, rows)
}
#[must_use]
pub fn palette_vec(&mut self) -> Vec<RGBA> {
let pal = self.palette();
let mut out: Vec<RGBA> = Vec::new();
out.try_reserve_exact(pal.len()).unwrap();
out.extend_from_slice(pal);
out
}
}
fn sort_palette(attr: &Attributes, palette: &mut PalF) {
let last_index_transparent = attr.last_index_transparent;
let mut tmp: ArrayVec<_, {MAX_COLORS}> = palette.iter_mut().map(|(c,p)| (*c, *p)).collect();
tmp.sort_by_key(|(color, pop)| {
let is_transparent = color.a <= MAX_TRANSP_A;
(is_transparent == last_index_transparent, Reverse(OrdFloat::<f32>::unchecked_new(pop.popularity())))
});
palette.iter_mut().zip(tmp).for_each(|((dcol, dpop), (scol, spop))| {
*dcol = scol;
*dpop = spop;
});
if last_index_transparent {
let alpha_index = palette.as_slice().iter().enumerate()
.filter(|(_, c)| c.a <= MAX_TRANSP_A)
.min_by_key(|(_, c)| OrdFloat::<f32>::unchecked_new(c.a))
.map(|(i, _)| i);
if let Some(alpha_index) = alpha_index {
let last_index = palette.as_slice().len() - 1;
palette.swap(last_index, alpha_index);
}
} else {
let num_transparent = palette.as_slice().iter().enumerate()
.filter(|(_, c)| c.a <= MAX_TRANSP_A)
.map(|(i, _)| i + 1) .max();
if let Some(num_transparent) = num_transparent {
attr.verbose_print(format!(" eliminated opaque tRNS-chunk entries...{} entr{} transparent", num_transparent, if num_transparent == 1 { "y" } else { "ies" }));
}
}
}
impl fmt::Debug for QuantizationResult {
#[cold]
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "QuantizationResult(q={})", self.quantization_quality().unwrap_or(0))
}
}
#[allow(clippy::or_fun_call)]
pub(crate) fn find_best_palette(attr: &Attributes, target_mse: f64, target_mse_is_zero: bool, max_mse: Option<f64>, mut hist: HistogramInternal, fixed_colors: &FixedColorsSet) -> Result<(PalF, Option<f64>), Error> {
let few_input_colors = hist.items.len() + fixed_colors.len() <= attr.max_colors as usize;
if few_input_colors && target_mse_is_zero {
return Ok(palette_from_histogram(&hist, attr.max_colors, fixed_colors));
}
let mut max_colors = attr.max_colors;
let total_trials = attr.feedback_loop_trials(hist.items.len()) as i16;
let mut trials_left = total_trials;
let mut best_palette = None;
let mut target_mse_overshoot = if total_trials > 0 { 1.05 } else { 1. };
let mut fails_in_a_row = 0;
let mut palette_error = None;
let mut palette = loop {
let max_mse_per_color = target_mse.max(palette_error.unwrap_or(quality_to_mse(1))).max(quality_to_mse(51)) * 1.2;
let mut new_palette = mediancut(&mut hist, max_colors - fixed_colors.len() as PalLen, target_mse * target_mse_overshoot, max_mse_per_color)?
.with_fixed_colors(max_colors, fixed_colors);
let stage_done = 1. - (trials_left.max(0) as f32 / (total_trials + 1) as f32).powi(2);
let overall_done = attr.progress_stage1 as f32 + stage_done * attr.progress_stage2 as f32;
attr.verbose_print(format!(" selecting colors...{}%", (100. * stage_done) as u8));
if trials_left <= 0 { break Some(new_palette); }
let first_run_of_target_mse = best_palette.is_none() && target_mse > 0.;
let total_error = Kmeans::iteration(&mut hist, &mut new_palette, !first_run_of_target_mse)?;
if best_palette.is_none() || total_error < palette_error.unwrap_or(f64::MAX) || (total_error <= target_mse && new_palette.len() < max_colors as usize) {
if total_error < target_mse && total_error > 0. {
target_mse_overshoot = if (target_mse_overshoot * 1.25) < (target_mse / total_error) {target_mse_overshoot * 1.25 } else {target_mse / total_error }; }
palette_error = Some(total_error);
max_colors = max_colors.min(new_palette.len() as PalLen + 1);
trials_left -= 1;
fails_in_a_row = 0;
best_palette = Some(new_palette);
} else {
fails_in_a_row += 1;
target_mse_overshoot = 1.;
trials_left -= 5 + fails_in_a_row;
}
if attr.progress(overall_done) || trials_left <= 0 {
break best_palette;
}
}.ok_or(ValueOutOfRange)?;
refine_palette(&mut palette, attr, &mut hist, max_mse, &mut palette_error)?;
Ok((palette, palette_error))
}
fn refine_palette(palette: &mut PalF, attr: &Attributes, hist: &mut HistogramInternal, max_mse: Option<f64>, palette_error: &mut Option<f64>) -> Result<(), Error> {
let (iterations, iteration_limit) = attr.kmeans_iterations(hist.items.len(), palette_error.is_some());
if iterations > 0 {
attr.verbose_print(" moving colormap towards local minimum");
let mut i = 0;
while i < iterations {
let stage_done = i as f32 / iterations as f32;
let overall_done = attr.progress_stage1 as f32 + attr.progress_stage2 as f32 + stage_done * attr.progress_stage3 as f32 * 0.89;
if attr.progress(overall_done) {
break;
}
let pal_err = Kmeans::iteration(hist, palette, false)?;
debug_assert!(pal_err < 1e20);
let previous_palette_error = *palette_error;
*palette_error = Some(pal_err);
if let Some(previous_palette_error) = previous_palette_error {
if (previous_palette_error - pal_err).abs() < iteration_limit {
break;
}
}
i += if pal_err > max_mse.unwrap_or(1e20) * 1.5 { 2 } else { 1 };
}
}
Ok(())
}
fn palette_from_histogram(hist: &HistogramInternal, max_colors: PalLen, fixed_colors: &FixedColorsSet) -> (PalF, Option<f64>) {
let mut hist_pal = PalF::new();
for item in hist.items.iter() {
hist_pal.push(item.color, PalPop::new(item.perceptual_weight));
}
(hist_pal.with_fixed_colors(max_colors, fixed_colors), Some(0.))
}
pub(crate) fn quality_to_mse(quality: u8) -> f64 {
if quality == 0 {
return 1e20; }
if quality >= 100 { return 0.; }
let extra_low_quality_fudge = (0.016 / (0.001 + quality as f64) - 0.001).max(0.);
LIQ_WEIGHT_MSE * (extra_low_quality_fudge + 2.5 / (210. + quality as f64).powf(1.2) * (100.1 - quality as f64) / 100.)
}
pub(crate) fn mse_to_quality(mse: f64) -> u8 {
for i in (1..101).rev() {
if mse <= quality_to_mse(i) + 0.000001 { return i; };
}
0
}