summaryrefslogtreecommitdiffstats
path: root/libavfilter/vf_owdenoise.c
blob: 3a77f8967e6d940f690a9e56b6fec13375d787af (plain)
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
/*
 * Copyright (c) 2007 Michael Niedermayer <michaelni@gmx.at>
 * Copyright (c) 2013 Clément Bœsch <u pkh me>
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License along
 * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 */

/**
 * @todo try to change to int
 * @todo try lifting based implementation
 * @todo optimize optimize optimize
 * @todo hard thresholding
 * @todo use QP to decide filter strength
 * @todo wavelet normalization / least squares optimal signal vs. noise thresholds
 */

#include "libavutil/imgutils.h"
#include "libavutil/opt.h"
#include "libavutil/pixdesc.h"
#include "avfilter.h"
#include "internal.h"

typedef struct {
    const AVClass *class;
    double luma_strength;
    double chroma_strength;
    int depth;
    float *plane[16+1][4];
    int linesize;
    int hsub, vsub;
} OWDenoiseContext;

#define OFFSET(x) offsetof(OWDenoiseContext, x)
#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
static const AVOption owdenoise_options[] = {
    { "depth",           "set depth",           OFFSET(depth),           AV_OPT_TYPE_INT,    {.i64 =   8}, 8,   16, FLAGS },
    { "luma_strength",   "set luma strength",   OFFSET(luma_strength),   AV_OPT_TYPE_DOUBLE, {.dbl = 1.0}, 0, 1000, FLAGS },
    { "ls",              "set luma strength",   OFFSET(luma_strength),   AV_OPT_TYPE_DOUBLE, {.dbl = 1.0}, 0, 1000, FLAGS },
    { "chroma_strength", "set chroma strength", OFFSET(chroma_strength), AV_OPT_TYPE_DOUBLE, {.dbl = 1.0}, 0, 1000, FLAGS },
    { "cs",              "set chroma strength", OFFSET(chroma_strength), AV_OPT_TYPE_DOUBLE, {.dbl = 1.0}, 0, 1000, FLAGS },
    { NULL }
};

AVFILTER_DEFINE_CLASS(owdenoise);

DECLARE_ALIGNED(8, static const uint8_t, dither)[8][8] = {
    {  0,  48,  12,  60,   3,  51,  15,  63 },
    { 32,  16,  44,  28,  35,  19,  47,  31 },
    {  8,  56,   4,  52,  11,  59,   7,  55 },
    { 40,  24,  36,  20,  43,  27,  39,  23 },
    {  2,  50,  14,  62,   1,  49,  13,  61 },
    { 34,  18,  46,  30,  33,  17,  45,  29 },
    { 10,  58,   6,  54,   9,  57,   5,  53 },
    { 42,  26,  38,  22,  41,  25,  37,  21 },
};

static const double coeff[2][5] = {
    {
         0.6029490182363579  * M_SQRT2,
         0.2668641184428723  * M_SQRT2,
        -0.07822326652898785 * M_SQRT2,
        -0.01686411844287495 * M_SQRT2,
         0.02674875741080976 * M_SQRT2,
    },{
         1.115087052456994   / M_SQRT2,
        -0.5912717631142470  / M_SQRT2,
        -0.05754352622849957 / M_SQRT2,
         0.09127176311424948 / M_SQRT2,
    }
};

static const double icoeff[2][5] = {
    {
         1.115087052456994   / M_SQRT2,
         0.5912717631142470  / M_SQRT2,
        -0.05754352622849957 / M_SQRT2,
        -0.09127176311424948 / M_SQRT2,
    },{
         0.6029490182363579  * M_SQRT2,
        -0.2668641184428723  * M_SQRT2,
        -0.07822326652898785 * M_SQRT2,
         0.01686411844287495 * M_SQRT2,
         0.02674875741080976 * M_SQRT2,
    }
};


static inline void decompose(float *dst_l, float *dst_h, const float *src,
                             int linesize, int w)
{
    int x, i;
    for (x = 0; x < w; x++) {
        double sum_l = src[x * linesize] * coeff[0][0];
        double sum_h = src[x * linesize] * coeff[1][0];
        for (i = 1; i <= 4; i++) {
            const double s = src[avpriv_mirror(x - i, w - 1) * linesize]
                           + src[avpriv_mirror(x + i, w - 1) * linesize];

            sum_l += coeff[0][i] * s;
            sum_h += coeff[1][i] * s;
        }
        dst_l[x * linesize] = sum_l;
        dst_h[x * linesize] = sum_h;
    }
}

static inline void compose(float *dst, const float *src_l, const float *src_h,
                           int linesize, int w)
{
    int x, i;
    for (x = 0; x < w; x++) {
        double sum_l = src_l[x * linesize] * icoeff[0][0];
        double sum_h = src_h[x * linesize] * icoeff[1][0];
        for (i = 1; i <= 4; i++) {
            const int x0 = avpriv_mirror(x - i, w - 1) * linesize;
            const int x1 = avpriv_mirror(x + i, w - 1) * linesize;

            sum_l += icoeff[0][i] * (src_l[x0] + src_l[x1]);
            sum_h += icoeff[1][i] * (src_h[x0] + src_h[x1]);
        }
        dst[x * linesize] = (sum_l + sum_h) * 0.5;
    }
}

static inline void decompose2D(float *dst_l, float *dst_h, const float *src,
                               int xlinesize, int ylinesize,
                               int step, int w, int h)
{
    int y, x;
    for (y = 0; y < h; y++)
        for (x = 0; x < step; x++)
            decompose(dst_l + ylinesize*y + xlinesize*x,
                      dst_h + ylinesize*y + xlinesize*x,
                      src   + ylinesize*y + xlinesize*x,
                      step * xlinesize, (w - x + step - 1) / step);
}

static inline void compose2D(float *dst, const float *src_l, const float *src_h,
                             int xlinesize, int ylinesize,
                             int step, int w, int h)
{
    int y, x;
    for (y = 0; y < h; y++)
        for (x = 0; x < step; x++)
            compose(dst   + ylinesize*y + xlinesize*x,
                    src_l + ylinesize*y + xlinesize*x,
                    src_h + ylinesize*y + xlinesize*x,
                    step * xlinesize, (w - x + step - 1) / step);
}

static void decompose2D2(float *dst[4], float *src, float *temp[2],
                         int linesize, int step, int w, int h)
{
    decompose2D(temp[0], temp[1], src,     1, linesize, step, w, h);
    decompose2D( dst[0],  dst[1], temp[0], linesize, 1, step, h, w);
    decompose2D( dst[2],  dst[3], temp[1], linesize, 1, step, h, w);
}

static void compose2D2(float *dst, float *src[4], float *temp[2],
                       int linesize, int step, int w, int h)
{
    compose2D(temp[0],  src[0],  src[1], linesize, 1, step, h, w);
    compose2D(temp[1],  src[2],  src[3], linesize, 1, step, h, w);
    compose2D(dst,     temp[0], temp[1], 1, linesize, step, w, h);
}

static void filter(OWDenoiseContext *s,
                   uint8_t       *dst, int dst_linesize,
                   const uint8_t *src, int src_linesize,
                   int width, int height, double strength)
{
    int x, y, i, j, depth = s->depth;

    while (1<<depth > width || 1<<depth > height)
        depth--;

    for (y = 0; y < height; y++)
        for(x = 0; x < width; x++)
            s->plane[0][0][y*s->linesize + x] = src[y*src_linesize + x];

    for (i = 0; i < depth; i++)
        decompose2D2(s->plane[i + 1], s->plane[i][0], s->plane[0] + 1, s->linesize, 1<<i, width, height);

    for (i = 0; i < depth; i++) {
        for (j = 1; j < 4; j++) {
            for (y = 0; y < height; y++) {
                for (x = 0; x < width; x++) {
                    double v = s->plane[i + 1][j][y*s->linesize + x];
                    if      (v >  strength) v -= strength;
                    else if (v < -strength) v += strength;
                    else                    v  = 0;
                    s->plane[i + 1][j][x + y*s->linesize] = v;
                }
            }
        }
    }
    for (i = depth-1; i >= 0; i--)
        compose2D2(s->plane[i][0], s->plane[i + 1], s->plane[0] + 1, s->linesize, 1<<i, width, height);

    for (y = 0; y < height; y++) {
        for (x = 0; x < width; x++) {
            i = s->plane[0][0][y*s->linesize + x] + dither[x&7][y&7]*(1.0/64) + 1.0/128; // yes the rounding is insane but optimal :)
            if ((unsigned)i > 255U) i = ~(i >> 31);
            dst[y*dst_linesize + x] = i;
        }
    }
}

static int filter_frame(AVFilterLink *inlink, AVFrame *in)
{
    int direct = 0;
    AVFilterContext *ctx = inlink->dst;
    OWDenoiseContext *s = ctx->priv;
    AVFilterLink *outlink = ctx->outputs[0];
    AVFrame *out;
    const int cw = AV_CEIL_RSHIFT(inlink->w, s->hsub);
    const int ch = AV_CEIL_RSHIFT(inlink->h, s->vsub);

    if (av_frame_is_writable(in)) {
        direct = 1;
        out = in;
    } else {
        out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
        if (!out) {
            av_frame_free(&in);
            return AVERROR(ENOMEM);
        }
        av_frame_copy_props(out, in);
    }

    filter(s, out->data[0], out->linesize[0], in->data[0], in->linesize[0], inlink->w, inlink->h, s->luma_strength);
    filter(s, out->data[1], out->linesize[1], in->data[1], in->linesize[1], cw,        ch,        s->chroma_strength);
    filter(s, out->data[2], out->linesize[2], in->data[2], in->linesize[2], cw,        ch,        s->chroma_strength);

    if (!direct) {
        if (in->data[3])
            av_image_copy_plane(out->data[3], out->linesize[3],
                                in ->data[3], in ->linesize[3],
                                inlink->w, inlink->h);
        av_frame_free(&in);
    }

    return ff_filter_frame(outlink, out);
}

static int query_formats(AVFilterContext *ctx)
{
    static const enum AVPixelFormat pix_fmts[] = {
        AV_PIX_FMT_YUV444P,      AV_PIX_FMT_YUV422P,
        AV_PIX_FMT_YUV420P,      AV_PIX_FMT_YUV411P,
        AV_PIX_FMT_YUV410P,      AV_PIX_FMT_YUV440P,
        AV_PIX_FMT_YUVA444P,     AV_PIX_FMT_YUVA422P,
        AV_PIX_FMT_YUVA420P,
        AV_PIX_FMT_NONE
    };
    AVFilterFormats *fmts_list = ff_make_format_list(pix_fmts);
    if (!fmts_list)
        return AVERROR(ENOMEM);
    return ff_set_common_formats(ctx, fmts_list);
}

static int config_input(AVFilterLink *inlink)
{
    int i, j;
    OWDenoiseContext *s = inlink->dst->priv;
    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
    const int h = FFALIGN(inlink->h, 16);

    s->hsub = desc->log2_chroma_w;
    s->vsub = desc->log2_chroma_h;

    s->linesize = FFALIGN(inlink->w, 16);
    for (j = 0; j < 4; j++) {
        for (i = 0; i <= s->depth; i++) {
            s->plane[i][j] = av_malloc_array(s->linesize, h * sizeof(s->plane[0][0][0]));
            if (!s->plane[i][j])
                return AVERROR(ENOMEM);
        }
    }
    return 0;
}

static av_cold void uninit(AVFilterContext *ctx)
{
    int i, j;
    OWDenoiseContext *s = ctx->priv;

    for (j = 0; j < 4; j++)
        for (i = 0; i <= s->depth; i++)
            av_freep(&s->plane[i][j]);
}

static const AVFilterPad owdenoise_inputs[] = {
    {
        .name         = "default",
        .type         = AVMEDIA_TYPE_VIDEO,
        .filter_frame = filter_frame,
        .config_props = config_input,
    },
    { NULL }
};

static const AVFilterPad owdenoise_outputs[] = {
    {
        .name = "default",
        .type = AVMEDIA_TYPE_VIDEO,
    },
    { NULL }
};

AVFilter ff_vf_owdenoise = {
    .name          = "owdenoise",
    .description   = NULL_IF_CONFIG_SMALL("Denoise using wavelets."),
    .priv_size     = sizeof(OWDenoiseContext),
    .uninit        = uninit,
    .query_formats = query_formats,
    .inputs        = owdenoise_inputs,
    .outputs       = owdenoise_outputs,
    .priv_class    = &owdenoise_class,
    .flags         = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
};
OpenPOWER on IntegriCloud