2 * Copyright (c) 2011 Pascal Getreuer
3 * Copyright (c) 2016 Paul B Mahol
5 * Redistribution and use in source and binary forms, with or without modification,
6 * are permitted provided that the following conditions are met:
8 * * Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * * Redistributions in binary form must reproduce the above
11 * copyright notice, this list of conditions and the following
12 * disclaimer in the documentation and/or other materials provided
13 * with the distribution.
15 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
16 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
17 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
18 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
19 * HOLDER BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
20 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
21 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
22 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
23 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
24 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
25 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30 #include "libavutil/imgutils.h"
31 #include "libavutil/opt.h"
32 #include "libavutil/pixdesc.h"
37 #include "vf_gblur_init.h"
40 #define OFFSET(x) offsetof(GBlurContext, x)
41 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
43 static const AVOption gblur_options[] = {
44 { "sigma", "set sigma", OFFSET(sigma), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, 0.0, 1024, FLAGS },
45 { "steps", "set number of steps", OFFSET(steps), AV_OPT_TYPE_INT, {.i64=1}, 1, 6, FLAGS },
46 { "planes", "set planes to filter", OFFSET(planes), AV_OPT_TYPE_INT, {.i64=0xF}, 0, 0xF, FLAGS },
47 { "sigmaV", "set vertical sigma", OFFSET(sigmaV), AV_OPT_TYPE_FLOAT, {.dbl=-1}, -1, 1024, FLAGS },
51 AVFILTER_DEFINE_CLASS(gblur);
53 typedef struct ThreadData {
58 static int filter_horizontally(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
60 GBlurContext *s = ctx->priv;
62 const int height = td->height;
63 const int width = td->width;
64 const int slice_start = (height * jobnr ) / nb_jobs;
65 const int slice_end = (height * (jobnr+1)) / nb_jobs;
66 const float boundaryscale = s->boundaryscale;
67 const int steps = s->steps;
68 const float nu = s->nu;
69 float *buffer = s->buffer;
70 float *localbuf = NULL;
73 localbuf = s->localbuf + s->stride * width * slice_start;
75 s->horiz_slice(buffer + width * slice_start, width, slice_end - slice_start,
76 steps, nu, boundaryscale, localbuf);
81 static int filter_vertically(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
83 GBlurContext *s = ctx->priv;
85 const int height = td->height;
86 const int width = td->width;
87 const int slice_start = (width * jobnr ) / nb_jobs;
88 const int slice_end = (width * (jobnr+1)) / nb_jobs;
89 const float boundaryscale = s->boundaryscaleV;
90 const int steps = s->steps;
91 const float nu = s->nuV;
92 float *buffer = s->buffer;
94 s->verti_slice(buffer, width, height, slice_start, slice_end,
95 steps, nu, boundaryscale);
100 static int filter_postscale(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
102 GBlurContext *s = ctx->priv;
103 ThreadData *td = arg;
104 const float max = s->flt ? FLT_MAX : (1 << s->depth) - 1;
105 const float min = s->flt ? -FLT_MAX : 0.f;
106 const int height = td->height;
107 const int width = td->width;
108 const int awidth = FFALIGN(width, 64);
109 const int slice_start = (height * jobnr ) / nb_jobs;
110 const int slice_end = (height * (jobnr+1)) / nb_jobs;
111 const float postscale = s->postscale * s->postscaleV;
112 const int slice_size = slice_end - slice_start;
114 s->postscale_slice(s->buffer + slice_start * awidth,
115 slice_size * awidth, postscale, min, max);
120 static void gaussianiir2d(AVFilterContext *ctx, int plane)
122 GBlurContext *s = ctx->priv;
123 const int width = s->planewidth[plane];
124 const int height = s->planeheight[plane];
125 const int nb_threads = ff_filter_get_nb_threads(ctx);
128 if (s->sigma <= 0 || s->steps < 0)
133 ff_filter_execute(ctx, filter_horizontally, &td,
134 NULL, FFMIN(height, nb_threads));
135 ff_filter_execute(ctx, filter_vertically, &td,
136 NULL, FFMIN(width, nb_threads));
137 ff_filter_execute(ctx, filter_postscale, &td,
138 NULL, FFMIN(width * height, nb_threads));
141 static const enum AVPixelFormat pix_fmts[] = {
142 AV_PIX_FMT_YUVA444P, AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV440P,
143 AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVJ440P,
144 AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUV420P,
145 AV_PIX_FMT_YUVJ422P, AV_PIX_FMT_YUVJ420P,
146 AV_PIX_FMT_YUVJ411P, AV_PIX_FMT_YUV411P, AV_PIX_FMT_YUV410P,
147 AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV444P9,
148 AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV444P10,
149 AV_PIX_FMT_YUV420P12, AV_PIX_FMT_YUV422P12, AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV440P12,
150 AV_PIX_FMT_YUV420P14, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV444P14,
151 AV_PIX_FMT_YUV420P16, AV_PIX_FMT_YUV422P16, AV_PIX_FMT_YUV444P16,
152 AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA444P9,
153 AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA444P10,
154 AV_PIX_FMT_YUVA422P12, AV_PIX_FMT_YUVA444P12,
155 AV_PIX_FMT_YUVA420P16, AV_PIX_FMT_YUVA422P16, AV_PIX_FMT_YUVA444P16,
156 AV_PIX_FMT_GBRP, AV_PIX_FMT_GBRP9, AV_PIX_FMT_GBRP10,
157 AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRP14, AV_PIX_FMT_GBRP16,
158 AV_PIX_FMT_GBRAP, AV_PIX_FMT_GBRAP10, AV_PIX_FMT_GBRAP12, AV_PIX_FMT_GBRAP16,
159 AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY9, AV_PIX_FMT_GRAY10, AV_PIX_FMT_GRAY12, AV_PIX_FMT_GRAY14, AV_PIX_FMT_GRAY16,
160 AV_PIX_FMT_GBRPF32, AV_PIX_FMT_GBRAPF32,
165 static av_cold void uninit(AVFilterContext *ctx)
167 GBlurContext *s = ctx->priv;
169 av_freep(&s->buffer);
170 av_freep(&s->localbuf);
173 static int config_input(AVFilterLink *inlink)
175 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
176 GBlurContext *s = inlink->dst->priv;
180 s->depth = desc->comp[0].depth;
181 s->flt = !!(desc->flags & AV_PIX_FMT_FLAG_FLOAT);
182 s->planewidth[1] = s->planewidth[2] = AV_CEIL_RSHIFT(inlink->w, desc->log2_chroma_w);
183 s->planewidth[0] = s->planewidth[3] = inlink->w;
184 s->planeheight[1] = s->planeheight[2] = AV_CEIL_RSHIFT(inlink->h, desc->log2_chroma_h);
185 s->planeheight[0] = s->planeheight[3] = inlink->h;
187 s->nb_planes = av_pix_fmt_count_planes(inlink->format);
189 s->buffer = av_malloc_array(FFALIGN(inlink->w, 64), FFALIGN(inlink->h, 64) * sizeof(*s->buffer));
191 return AVERROR(ENOMEM);
194 s->sigmaV = s->sigma;
201 static void set_params(float sigma, int steps, float *postscale, float *boundaryscale, float *nu)
205 lambda = (sigma * sigma) / (2.0 * steps);
206 dnu = (1.0 + 2.0 * lambda - sqrt(1.0 + 4.0 * lambda)) / (2.0 * lambda);
207 *postscale = pow(dnu / lambda, steps);
208 *boundaryscale = 1.0 / (1.0 - dnu);
212 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
214 AVFilterContext *ctx = inlink->dst;
215 GBlurContext *s = ctx->priv;
216 AVFilterLink *outlink = ctx->outputs[0];
220 set_params(s->sigma, s->steps, &s->postscale, &s->boundaryscale, &s->nu);
221 set_params(s->sigmaV, s->steps, &s->postscaleV, &s->boundaryscaleV, &s->nuV);
223 if (av_frame_is_writable(in)) {
226 out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
229 return AVERROR(ENOMEM);
231 av_frame_copy_props(out, in);
234 for (plane = 0; plane < s->nb_planes; plane++) {
235 const int height = s->planeheight[plane];
236 const int width = s->planewidth[plane];
237 float *bptr = s->buffer;
238 const uint8_t *src = in->data[plane];
239 const uint16_t *src16 = (const uint16_t *)in->data[plane];
240 uint8_t *dst = out->data[plane];
241 uint16_t *dst16 = (uint16_t *)out->data[plane];
244 if (!s->sigma || !(s->planes & (1 << plane))) {
246 av_image_copy_plane(out->data[plane], out->linesize[plane],
247 in->data[plane], in->linesize[plane],
248 width * ((s->depth + 7) / 8), height);
253 av_image_copy_plane((uint8_t *)bptr, width * sizeof(float),
254 in->data[plane], in->linesize[plane],
255 width * sizeof(float), height);
256 } else if (s->depth == 8) {
257 for (y = 0; y < height; y++) {
258 for (x = 0; x < width; x++) {
262 src += in->linesize[plane];
265 for (y = 0; y < height; y++) {
266 for (x = 0; x < width; x++) {
270 src16 += in->linesize[plane] / 2;
274 gaussianiir2d(ctx, plane);
278 av_image_copy_plane(out->data[plane], out->linesize[plane],
279 (uint8_t *)bptr, width * sizeof(float),
280 width * sizeof(float), height);
281 } else if (s->depth == 8) {
282 for (y = 0; y < height; y++) {
283 for (x = 0; x < width; x++)
284 dst[x] = lrintf(bptr[x]);
286 dst += out->linesize[plane];
289 for (y = 0; y < height; y++) {
290 for (x = 0; x < width; x++)
291 dst16[x] = lrintf(bptr[x]);
293 dst16 += out->linesize[plane] / 2;
300 return ff_filter_frame(outlink, out);
303 static const AVFilterPad gblur_inputs[] = {
306 .type = AVMEDIA_TYPE_VIDEO,
307 .config_props = config_input,
308 .filter_frame = filter_frame,
312 static const AVFilterPad gblur_outputs[] = {
315 .type = AVMEDIA_TYPE_VIDEO,
319 const AVFilter ff_vf_gblur = {
321 .description = NULL_IF_CONFIG_SMALL("Apply Gaussian Blur filter."),
322 .priv_size = sizeof(GBlurContext),
323 .priv_class = &gblur_class,
325 FILTER_INPUTS(gblur_inputs),
326 FILTER_OUTPUTS(gblur_outputs),
327 FILTER_PIXFMTS_ARRAY(pix_fmts),
328 .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC | AVFILTER_FLAG_SLICE_THREADS,
329 .process_command = ff_filter_process_command,