February 2020 - ffmpeg-cvslog

avfilter/vf_v360: improve interpolation for equirect input at poles
by Paul B Mahol 29 Feb '20

29 Feb '20

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Sat Feb 29 22:27:11 2020 +0100| [8f3df1dd4d0eb8cdc59af9cb77b213630bf6af8b] | committer: Paul B Mahol avfilter/vf_v360: improve interpolation for equirect input at poles > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=8f3df1dd4d0eb8cdc… --- libavfilter/vf_v360.c | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-) diff --git a/libavfilter/vf_v360.c b/libavfilter/vf_v360.c index fc799f18c2..b6195a01f6 100644 --- a/libavfilter/vf_v360.c +++ b/libavfilter/vf_v360.c @@ -639,6 +639,22 @@ static inline int reflecty(int y, int h) return y; } +/** + * Reflect x operation for equirect. + * + * @param x input horizontal position + * @param y input vertical position + * @param w input width + * @param h input height + */ +static inline int ereflectx(int x, int y, int w, int h) +{ + if (y < 0 || y >= h) + x += w / 2; + + return mod(x, w); +} + /** * Reflect x operation. * @@ -1745,8 +1761,8 @@ static int xyz_to_equirect(const V360Context *s, for (int i = 0; i < 4; i++) { for (int j = 0; j < 4; j++) { - us[i][j] = mod(ui + j - 1, width); - vs[i][j] = av_clip(vi + i - 1, 0, height - 1); + us[i][j] = ereflectx(ui + j - 1, vi + i - 1, width, height); + vs[i][j] = reflecty(vi + i - 1, height); } }

1 0

avfilter/vf_cas: Remove superfluous ;
by Andreas Rheinhardt 29 Feb '20

29 Feb '20

ffmpeg | branch: master | Andreas Rheinhardt <andreas.rheinhardt(a)gmail.com> | Fri Feb 28 22:06:30 2020 +0100| [be82dc175bb8386c3fafc1ea9814bacdb8126eb0] | committer: Michael Niedermayer avfilter/vf_cas: Remove superfluous ; The second ; in a double ;; is actually a null statement. It triggers the typical declaration-after-statement compiler-warnings if it occurs in the middle of several declarations (like here). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt(a)gmail.com> Reviewed-by: Paul B Mahol <onemda(a)gmail.com> Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=be82dc175bb8386c3… --- libavfilter/vf_cas.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/libavfilter/vf_cas.c b/libavfilter/vf_cas.c index a07d51a84e..7c32ed5f9b 100644 --- a/libavfilter/vf_cas.c +++ b/libavfilter/vf_cas.c @@ -58,7 +58,7 @@ static int cas_slice8(AVFilterContext *avctx, void *arg, int jobnr, int nb_jobs) const int linesize = out->linesize[p]; const int in_linesize = in->linesize[p]; const int w = s->planewidth[p]; - const int w1 = w - 1;; + const int w1 = w - 1; const int h = s->planeheight[p]; const int h1 = h - 1; uint8_t *dst = out->data[p] + slice_start * linesize; @@ -125,7 +125,7 @@ static int cas_slice16(AVFilterContext *avctx, void *arg, int jobnr, int nb_jobs const int linesize = out->linesize[p] / 2; const int in_linesize = in->linesize[p] / 2; const int w = s->planewidth[p]; - const int w1 = w - 1;; + const int w1 = w - 1; const int h = s->planeheight[p]; const int h1 = h - 1; uint16_t *dst = (uint16_t *)out->data[p] + slice_start * linesize;

1 0

fftools/ffmpeg_opt: Fix leak of options when parsing options fails
by Andreas Rheinhardt 29 Feb '20

29 Feb '20

ffmpeg | branch: master | Andreas Rheinhardt <andreas.rheinhardt(a)gmail.com> | Fri Feb 28 22:06:29 2020 +0100| [21265f42ecb265debe9fec1dbfd0cb7de5a8aefb] | committer: Michael Niedermayer fftools/ffmpeg_opt: Fix leak of options when parsing options fails Fixes #8094. Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt(a)gmail.com> Reviewed-by: Paul B Mahol <onemda(a)gmail.com> Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=21265f42ecb265deb… --- fftools/ffmpeg_opt.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fftools/ffmpeg_opt.c b/fftools/ffmpeg_opt.c index 12d44886ee..3bf90bd20c 100644 --- a/fftools/ffmpeg_opt.c +++ b/fftools/ffmpeg_opt.c @@ -3278,6 +3278,7 @@ static int open_files(OptionGroupList *l, const char *inout, if (ret < 0) { av_log(NULL, AV_LOG_ERROR, "Error parsing options for %s file " "%s.\n", inout, g->arg); + uninit_options(&o); return ret; }

1 0

avfilter/vf_v360: add partial size setup for flat
by Paul B Mahol 29 Feb '20

29 Feb '20

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Sat Feb 29 20:22:37 2020 +0100| [f707c84b8b89a937cc388e5ae4b8fcd67e07b75c] | committer: Paul B Mahol avfilter/vf_v360: add partial size setup for flat Other part of size is calculated from both available horizontal and vertical FOV and given one size component. > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=f707c84b8b89a937c… --- libavfilter/vf_v360.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/libavfilter/vf_v360.c b/libavfilter/vf_v360.c index 3c73ad32f7..fc799f18c2 100644 --- a/libavfilter/vf_v360.c +++ b/libavfilter/vf_v360.c @@ -3790,7 +3790,15 @@ static int config_output(AVFilterLink *outlink) } // Override resolution with user values if specified - if (s->width > 0 && s->height > 0) { + if (s->width > 0 && s->height <= 0 && s->h_fov > 0.f && s->v_fov > 0.f && + s->out == FLAT && s->d_fov == 0.f) { + w = s->width; + h = w / tanf(s->h_fov * M_PI / 360.f) * tanf(s->v_fov * M_PI / 360.f); + } else if (s->width <= 0 && s->height > 0 && s->h_fov > 0.f && s->v_fov > 0.f && + s->out == FLAT && s->d_fov == 0.f) { + h = s->height; + w = h / tanf(s->v_fov * M_PI / 360.f) * tanf(s->h_fov * M_PI / 360.f); + } else if (s->width > 0 && s->height > 0) { w = s->width; h = s->height; } else if (s->width > 0 || s->height > 0) {

1 0

avfilter/vf_v360: speed up fisheye input calculation
by Paul B Mahol 29 Feb '20

29 Feb '20

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Sat Feb 29 16:39:35 2020 +0100| [3733a6bc20e2a2ca7b06f5d867542f1f5cc874f1] | committer: Paul B Mahol avfilter/vf_v360: speed up fisheye input calculation > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=3733a6bc20e2a2ca7… --- libavfilter/vf_v360.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/libavfilter/vf_v360.c b/libavfilter/vf_v360.c index 90e45b4fd0..3c73ad32f7 100644 --- a/libavfilter/vf_v360.c +++ b/libavfilter/vf_v360.c @@ -2484,11 +2484,12 @@ static int xyz_to_fisheye(const V360Context *s, const float *vec, int width, int height, int16_t us[4][4], int16_t vs[4][4], float *du, float *dv) { - const float phi = -atan2f(hypotf(vec[0], vec[1]), -vec[2]) / M_PI; - const float theta = -atan2f(vec[0], vec[1]); + const float h = hypotf(vec[0], vec[1]); + const float lh = h > 0.f ? h : 1.f; + const float phi = atan2f(h, -vec[2]) / M_PI; - float uf = sinf(theta) * phi * s->input_mirror_modifier[0] / s->iflat_range[0]; - float vf = cosf(theta) * phi * s->input_mirror_modifier[1] / s->iflat_range[1]; + float uf = vec[0] / lh * phi * s->input_mirror_modifier[0] / s->iflat_range[0]; + float vf = -vec[1] / lh * phi * s->input_mirror_modifier[1] / s->iflat_range[1]; const int visible = hypotf(uf, vf) <= 0.5f; int ui, vi;

1 0

avfilter/vf_v360: improve tetrahedron input format at frame borders
by Paul B Mahol 29 Feb '20

29 Feb '20

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Sat Feb 29 14:22:51 2020 +0100| [2dfd9445ff9b69a2b826e7b4c21c316689000147] | committer: Paul B Mahol avfilter/vf_v360: improve tetrahedron input format at frame borders > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=2dfd9445ff9b69a2b… --- libavfilter/vf_v360.c | 37 +++++++++++++++++++++++++++++++++++-- 1 file changed, 35 insertions(+), 2 deletions(-) diff --git a/libavfilter/vf_v360.c b/libavfilter/vf_v360.c index 1cdbaf7330..90e45b4fd0 100644 --- a/libavfilter/vf_v360.c +++ b/libavfilter/vf_v360.c @@ -622,6 +622,39 @@ static inline int mod(int a, int b) } } +/** + * Reflect y operation. + * + * @param y input vertical position + * @param h input height + */ +static inline int reflecty(int y, int h) +{ + if (y < 0) { + return -y; + } else if (y >= h) { + return 2 * h - 1 - y; + } + + return y; +} + +/** + * Reflect x operation. + * + * @param x input horizontal position + * @param y input vertical position + * @param w input width + * @param h input height + */ +static inline int reflectx(int x, int y, int w, int h) +{ + if (y < 0 || y >= h) + return w - 1 - x; + + return mod(x, w); +} + /** * Convert char to corresponding direction. * Used for cubemap options. @@ -2746,8 +2779,8 @@ static int xyz_to_tetrahedron(const V360Context *s, for (int i = 0; i < 4; i++) { for (int j = 0; j < 4; j++) { - us[i][j] = mod(ui + j - 1, width); - vs[i][j] = av_clip(vi + i - 1, 0, height - 1); + us[i][j] = reflectx(ui + j - 1, vi + i - 1, width, height); + vs[i][j] = reflecty(vi + i - 1, height); } }

1 0

avfilter/vf_v360: simplify tetrahedron input calculation
by Paul B Mahol 29 Feb '20

29 Feb '20

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Sat Feb 29 13:00:25 2020 +0100| [50a13b987b609f3280d2bc57cacbb67e2cf80521] | committer: Paul B Mahol avfilter/vf_v360: simplify tetrahedron input calculation > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=50a13b987b609f328… --- libavfilter/vf_v360.c | 13 +++++-------- 1 file changed, 5 insertions(+), 8 deletions(-) diff --git a/libavfilter/vf_v360.c b/libavfilter/vf_v360.c index 31d5e60553..1cdbaf7330 100644 --- a/libavfilter/vf_v360.c +++ b/libavfilter/vf_v360.c @@ -2713,18 +2713,15 @@ static int xyz_to_tetrahedron(const V360Context *s, const float *vec, int width, int height, int16_t us[4][4], int16_t vs[4][4], float *du, float *dv) { - float d = 0.5f * (vec[0] * vec[0] + vec[1] * vec[1] + vec[2] * vec[2]); - - const float d0 = (vec[0] * 0.5f + vec[1] * 0.5f + vec[2] *-0.5f) / d; - const float d1 = (vec[0] *-0.5f + vec[1] *-0.5f + vec[2] *-0.5f) / d; - const float d2 = (vec[0] * 0.5f + vec[1] *-0.5f + vec[2] * 0.5f) / d; - const float d3 = (vec[0] *-0.5f + vec[1] * 0.5f + vec[2] * 0.5f) / d; + const float d0 = vec[0] * 1.f + vec[1] * 1.f + vec[2] *-1.f; + const float d1 = vec[0] *-1.f + vec[1] *-1.f + vec[2] *-1.f; + const float d2 = vec[0] * 1.f + vec[1] *-1.f + vec[2] * 1.f; + const float d3 = vec[0] *-1.f + vec[1] * 1.f + vec[2] * 1.f; + const float d = FFMAX(d0, FFMAX3(d1, d2, d3)); float uf, vf, x, y, z; int ui, vi; - d = FFMAX(d0, FFMAX3(d1, d2, d3)); - x = vec[0] / d; y = vec[1] / d; z = -vec[2] / d;

1 0

avfilter/vf_v360: improve precision of some output formats
by Paul B Mahol 29 Feb '20

29 Feb '20

ffmpeg | branch: master | Paul B Mahol <onemda(a)gmail.com> | Sat Feb 29 12:59:23 2020 +0100| [fc7d4d08fb9f8059aaba8d2b1e1a89400faa50a2] | committer: Paul B Mahol avfilter/vf_v360: improve precision of some output formats > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=fc7d4d08fb9f8059a… --- libavfilter/vf_v360.c | 36 ++++++++++++++++++------------------ 1 file changed, 18 insertions(+), 18 deletions(-) diff --git a/libavfilter/vf_v360.c b/libavfilter/vf_v360.c index 068ba6dfce..31d5e60553 100644 --- a/libavfilter/vf_v360.c +++ b/libavfilter/vf_v360.c @@ -1610,8 +1610,8 @@ static int stereographic_to_xyz(const V360Context *s, int i, int j, int width, int height, float *vec) { - const float x = ((2.f * i) / width - 1.f) * s->flat_range[0]; - const float y = ((2.f * j) / height - 1.f) * s->flat_range[1]; + const float x = ((2.f * i + 1.f) / width - 1.f) * s->flat_range[0]; + const float y = ((2.f * j + 1.f) / height - 1.f) * s->flat_range[1]; const float xy = x * x + y * y; vec[0] = 2.f * x / (1.f + xy); @@ -1837,8 +1837,8 @@ static int mercator_to_xyz(const V360Context *s, int i, int j, int width, int height, float *vec) { - const float phi = ((2.f * i) / width - 1.f) * M_PI + M_PI_2; - const float y = ((2.f * j) / height - 1.f) * M_PI; + const float phi = ((2.f * i + 1.f) / width - 1.f) * M_PI + M_PI_2; + const float y = ((2.f * j + 1.f) / height - 1.f) * M_PI; const float div = expf(2.f * y) + 1.f; const float sin_phi = sinf(phi); @@ -1907,8 +1907,8 @@ static int ball_to_xyz(const V360Context *s, int i, int j, int width, int height, float *vec) { - const float x = (2.f * i) / width - 1.f; - const float y = (2.f * j) / height - 1.f; + const float x = (2.f * i + 1.f) / width - 1.f; + const float y = (2.f * j + 1.f) / height - 1.f; const float l = hypotf(x, y); if (l <= 1.f) { @@ -1941,8 +1941,8 @@ static int hammer_to_xyz(const V360Context *s, int i, int j, int width, int height, float *vec) { - const float x = ((2.f * i) / width - 1.f); - const float y = ((2.f * j) / height - 1.f); + const float x = ((2.f * i + 1.f) / width - 1.f); + const float y = ((2.f * j + 1.f) / height - 1.f); const float xx = x * x; const float yy = y * y; @@ -2022,8 +2022,8 @@ static int sinusoidal_to_xyz(const V360Context *s, int i, int j, int width, int height, float *vec) { - const float theta = ((2.f * j) / height - 1.f) * M_PI_2; - const float phi = ((2.f * i) / width - 1.f) * M_PI / cosf(theta); + const float theta = ((2.f * j + 1.f) / height - 1.f) * M_PI_2; + const float phi = ((2.f * i + 1.f) / width - 1.f) * M_PI / cosf(theta); const float sin_phi = sinf(phi); const float cos_phi = cosf(phi); @@ -2360,8 +2360,8 @@ static int flat_to_xyz(const V360Context *s, int i, int j, int width, int height, float *vec) { - const float l_x = s->flat_range[0] * (2.f * i / width - 1.f); - const float l_y = -s->flat_range[1] * (2.f * j / height - 1.f); + const float l_x = s->flat_range[0] * ((2.f * i + 0.5f) / width - 1.f); + const float l_y = -s->flat_range[1] * ((2.f * j + 0.5f) / height - 1.f); vec[0] = l_x; vec[1] = l_y; @@ -2493,8 +2493,8 @@ static int pannini_to_xyz(const V360Context *s, int i, int j, int width, int height, float *vec) { - const float uf = ((2.f * i) / width - 1.f); - const float vf = ((2.f * j) / height - 1.f); + const float uf = ((2.f * i + 1.f) / width - 1.f); + const float vf = ((2.f * j + 1.f) / height - 1.f); const float d = s->h_fov; const float k = uf * uf / ((d + 1.f) * (d + 1.f)); @@ -2544,8 +2544,8 @@ static int cylindrical_to_xyz(const V360Context *s, int i, int j, int width, int height, float *vec) { - const float uf = s->flat_range[0] * ((2.f * i) / width - 1.f); - const float vf = s->flat_range[1] * ((2.f * j) / height - 1.f); + const float uf = s->flat_range[0] * ((2.f * i + 1.f) / width - 1.f); + const float vf = s->flat_range[1] * ((2.f * j + 1.f) / height - 1.f); const float phi = uf; const float theta = atanf(vf); @@ -2638,8 +2638,8 @@ static int perspective_to_xyz(const V360Context *s, int i, int j, int width, int height, float *vec) { - const float uf = ((2.f * i) / width - 1.f); - const float vf = ((2.f * j) / height - 1.f); + const float uf = ((2.f * i + 1.f) / width - 1.f); + const float vf = ((2.f * j + 1.f) / height - 1.f); const float rh = hypotf(uf, vf); const float sinzz = 1.f - rh * rh; const float h = 1.f + s->v_fov;

1 0

avcodec/cdtoons: Remove superfluous ;
by Andreas Rheinhardt 28 Feb '20

28 Feb '20

ffmpeg | branch: master | Andreas Rheinhardt <andreas.rheinhardt(a)gmail.com> | Thu Feb 27 06:33:51 2020 +0100| [b6879b61dfe69e79c0fd32ce076fe1e7cbcc9c0c] | committer: Michael Niedermayer avcodec/cdtoons: Remove superfluous ; The second ; in a double ;; is actually a null statement. It triggers the typical declaration-after-statement compiler-warnings if it occurs in the middle of several declarations (like here). Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt(a)gmail.com> Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=b6879b61dfe69e79c… --- libavcodec/cdtoons.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libavcodec/cdtoons.c b/libavcodec/cdtoons.c index d5dce6351f..13f9a60f0f 100644 --- a/libavcodec/cdtoons.c +++ b/libavcodec/cdtoons.c @@ -61,7 +61,7 @@ static int cdtoons_render_sprite(AVCodecContext *avctx, const uint8_t *data, { CDToonsContext *c = avctx->priv_data; const uint8_t *next_line = data; - const uint8_t *end = data + data_size;; + const uint8_t *end = data + data_size; uint16_t line_size; uint8_t *dest; int skip = 0, to_skip, x;

1 0

avformat/mp3dec: Count last partial frame in probe.
by Michael Niedermayer 28 Feb '20

28 Feb '20

ffmpeg | branch: master | Michael Niedermayer <michael(a)niedermayer.cc> | Wed Feb 26 18:38:33 2020 +0100| [23645f95ec0ae3f4ccc8ad2b23d1fbb16575ae16] | committer: Michael Niedermayer avformat/mp3dec: Count last partial frame in probe. Fixes: regression Fixes: Ticket8511 Reviewed-by: Anton Khirnov <anton(a)khirnov.net> Signed-off-by: Michael Niedermayer <michael(a)niedermayer.cc> > http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=23645f95ec0ae3f4c… --- libavformat/mp3dec.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) diff --git a/libavformat/mp3dec.c b/libavformat/mp3dec.c index 71a4ed706d..b044679c02 100644 --- a/libavformat/mp3dec.c +++ b/libavformat/mp3dec.c @@ -87,20 +87,26 @@ static int mp3_read_probe(const AVProbeData *p) for (framesizes = frames = 0; buf2 < end; frames++) { MPADecodeHeader h; int header_emu = 0; + int available; header = AV_RB32(buf2); ret = avpriv_mpegaudio_decode_header(&h, header); - if (ret != 0 || end - buf2 < h.frame_size) + if (ret != 0) break; - for (buf3 = buf2 + 4; buf3 < buf2 + h.frame_size; buf3++) { + available = FFMIN(h.frame_size, end - buf2); + for (buf3 = buf2 + 4; buf3 < buf2 + available; buf3++) { uint32_t next_sync = AV_RB32(buf3); header_emu += (next_sync & MP3_MASK) == (header & MP3_MASK); } if (header_emu > 2) break; - buf2 += h.frame_size; framesizes += h.frame_size; + if (available < h.frame_size) { + frames++; + break; + } + buf2 += h.frame_size; } max_frames = FFMAX(max_frames, frames); max_framesizes = FFMAX(max_framesizes, framesizes);

1 0