2009-07-07 02:01:36 +00:00
|
|
|
/*
|
|
|
|
* Copyright (C) 2009 Loren Merritt <lorenm@u.washignton.edu>
|
|
|
|
*
|
|
|
|
* This file is part of MPlayer.
|
|
|
|
*
|
|
|
|
* MPlayer 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.
|
|
|
|
*
|
|
|
|
* MPlayer 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 MPlayer; if not, write to the Free Software Foundation, Inc.,
|
|
|
|
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Debanding algorithm (from gradfun2db by prunedtree):
|
|
|
|
* Boxblur.
|
|
|
|
* Foreach pixel, if it's within threshold of the blurred value, make it closer.
|
|
|
|
* So now we have a smoothed and higher bitdepth version of all the shallow
|
|
|
|
* gradients, while leaving detailed areas untouched.
|
|
|
|
* Dither it back to 8bit.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <stdio.h>
|
|
|
|
#include <stdlib.h>
|
|
|
|
#include <string.h>
|
|
|
|
#include <inttypes.h>
|
2012-01-08 17:25:26 +00:00
|
|
|
#include <math.h>
|
2009-07-07 02:01:36 +00:00
|
|
|
|
2012-08-15 20:23:02 +00:00
|
|
|
#include <libavutil/common.h>
|
|
|
|
|
2009-07-07 02:01:36 +00:00
|
|
|
#include "config.h"
|
2012-11-09 00:06:43 +00:00
|
|
|
#include "core/cpudetect.h"
|
|
|
|
#include "video/img_format.h"
|
|
|
|
#include "video/mp_image.h"
|
2009-07-07 02:01:36 +00:00
|
|
|
#include "vf.h"
|
2012-11-09 00:06:43 +00:00
|
|
|
#include "video/memcpy_pic.h"
|
|
|
|
#include "compat/x86_cpu.h"
|
2009-07-07 02:01:36 +00:00
|
|
|
|
2012-11-09 00:06:43 +00:00
|
|
|
#include "core/m_option.h"
|
|
|
|
#include "core/m_struct.h"
|
2012-01-08 17:12:43 +00:00
|
|
|
|
2009-07-07 02:01:36 +00:00
|
|
|
struct vf_priv_s {
|
2012-01-08 17:12:43 +00:00
|
|
|
float cfg_thresh;
|
|
|
|
int cfg_radius;
|
2012-01-08 17:25:26 +00:00
|
|
|
float cfg_size;
|
2009-07-07 02:01:36 +00:00
|
|
|
int thresh;
|
|
|
|
int radius;
|
|
|
|
uint16_t *buf;
|
|
|
|
void (*filter_line)(uint8_t *dst, uint8_t *src, uint16_t *dc,
|
|
|
|
int width, int thresh, const uint16_t *dithers);
|
|
|
|
void (*blur_line)(uint16_t *dc, uint16_t *buf, uint16_t *buf1,
|
|
|
|
uint8_t *src, int sstride, int width);
|
2012-01-08 17:12:43 +00:00
|
|
|
} const vf_priv_dflt = {
|
2012-01-10 20:18:26 +00:00
|
|
|
.cfg_thresh = 1.5,
|
2012-01-08 17:25:26 +00:00
|
|
|
.cfg_radius = -1,
|
|
|
|
.cfg_size = -1,
|
2009-07-07 02:01:36 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
static const uint16_t __attribute__((aligned(16))) pw_7f[8] = {127,127,127,127,127,127,127,127};
|
|
|
|
static const uint16_t __attribute__((aligned(16))) pw_ff[8] = {255,255,255,255,255,255,255,255};
|
|
|
|
static const uint16_t __attribute__((aligned(16))) dither[8][8] = {
|
|
|
|
{ 0, 96, 24,120, 6,102, 30,126 },
|
|
|
|
{ 64, 32, 88, 56, 70, 38, 94, 62 },
|
|
|
|
{ 16,112, 8,104, 22,118, 14,110 },
|
|
|
|
{ 80, 48, 72, 40, 86, 54, 78, 46 },
|
|
|
|
{ 4,100, 28,124, 2, 98, 26,122 },
|
|
|
|
{ 68, 36, 92, 60, 66, 34, 90, 58 },
|
|
|
|
{ 20,116, 12,108, 18,114, 10,106 },
|
|
|
|
{ 84, 52, 76, 44, 82, 50, 74, 42 },
|
|
|
|
};
|
|
|
|
|
|
|
|
static void filter_line_c(uint8_t *dst, uint8_t *src, uint16_t *dc,
|
|
|
|
int width, int thresh, const uint16_t *dithers)
|
|
|
|
{
|
|
|
|
int x;
|
|
|
|
for (x=0; x<width; x++, dc+=x&1) {
|
|
|
|
int pix = src[x]<<7;
|
|
|
|
int delta = dc[0] - pix;
|
|
|
|
int m = abs(delta) * thresh >> 16;
|
|
|
|
m = FFMAX(0, 127-m);
|
|
|
|
m = m*m*delta >> 14;
|
|
|
|
pix += m + dithers[x&7];
|
|
|
|
dst[x] = av_clip_uint8(pix>>7);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void blur_line_c(uint16_t *dc, uint16_t *buf, uint16_t *buf1,
|
|
|
|
uint8_t *src, int sstride, int width)
|
|
|
|
{
|
|
|
|
int x, v, old;
|
|
|
|
for (x=0; x<width; x++) {
|
|
|
|
v = buf1[x] + src[2*x] + src[2*x+1] + src[2*x+sstride] + src[2*x+1+sstride];
|
|
|
|
old = buf[x];
|
|
|
|
buf[x] = v;
|
|
|
|
dc[x] = v - old;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2010-04-05 16:38:40 +00:00
|
|
|
#if HAVE_MMX2
|
2009-07-07 02:01:36 +00:00
|
|
|
static void filter_line_mmx2(uint8_t *dst, uint8_t *src, uint16_t *dc,
|
|
|
|
int width, int thresh, const uint16_t *dithers)
|
|
|
|
{
|
|
|
|
intptr_t x;
|
|
|
|
if (width&3) {
|
|
|
|
x = width&~3;
|
|
|
|
filter_line_c(dst+x, src+x, dc+x/2, width-x, thresh, dithers);
|
|
|
|
width = x;
|
|
|
|
}
|
|
|
|
x = -width;
|
2010-03-26 08:13:10 +00:00
|
|
|
__asm__ volatile(
|
2009-07-07 02:01:36 +00:00
|
|
|
"movd %4, %%mm5 \n"
|
|
|
|
"pxor %%mm7, %%mm7 \n"
|
|
|
|
"pshufw $0, %%mm5, %%mm5 \n"
|
|
|
|
"movq %6, %%mm6 \n"
|
|
|
|
"movq %5, %%mm4 \n"
|
|
|
|
"1: \n"
|
|
|
|
"movd (%2,%0), %%mm0 \n"
|
|
|
|
"movd (%3,%0), %%mm1 \n"
|
|
|
|
"punpcklbw %%mm7, %%mm0 \n"
|
|
|
|
"punpcklwd %%mm1, %%mm1 \n"
|
|
|
|
"psllw $7, %%mm0 \n"
|
|
|
|
"pxor %%mm2, %%mm2 \n"
|
|
|
|
"psubw %%mm0, %%mm1 \n" // delta = dc - pix
|
|
|
|
"psubw %%mm1, %%mm2 \n"
|
|
|
|
"pmaxsw %%mm1, %%mm2 \n"
|
|
|
|
"pmulhuw %%mm5, %%mm2 \n" // m = abs(delta) * thresh >> 16
|
|
|
|
"psubw %%mm6, %%mm2 \n"
|
|
|
|
"pminsw %%mm7, %%mm2 \n" // m = -max(0, 127-m)
|
|
|
|
"pmullw %%mm2, %%mm2 \n"
|
|
|
|
"paddw %%mm4, %%mm0 \n" // pix += dither
|
|
|
|
"pmulhw %%mm2, %%mm1 \n"
|
|
|
|
"psllw $2, %%mm1 \n" // m = m*m*delta >> 14
|
|
|
|
"paddw %%mm1, %%mm0 \n" // pix += m
|
|
|
|
"psraw $7, %%mm0 \n"
|
|
|
|
"packuswb %%mm0, %%mm0 \n"
|
|
|
|
"movd %%mm0, (%1,%0) \n" // dst = clip(pix>>7)
|
|
|
|
"add $4, %0 \n"
|
|
|
|
"jl 1b \n"
|
|
|
|
"emms \n"
|
|
|
|
:"+r"(x)
|
|
|
|
:"r"(dst+width), "r"(src+width), "r"(dc+width/2),
|
|
|
|
"rm"(thresh), "m"(*dithers), "m"(*pw_7f)
|
|
|
|
:"memory"
|
|
|
|
);
|
|
|
|
}
|
2010-04-05 16:38:40 +00:00
|
|
|
#endif
|
2009-07-07 02:01:36 +00:00
|
|
|
|
2010-04-05 16:38:40 +00:00
|
|
|
#if HAVE_SSSE3
|
2009-07-07 02:01:36 +00:00
|
|
|
static void filter_line_ssse3(uint8_t *dst, uint8_t *src, uint16_t *dc,
|
|
|
|
int width, int thresh, const uint16_t *dithers)
|
|
|
|
{
|
|
|
|
intptr_t x;
|
|
|
|
if (width&7) {
|
|
|
|
// could be 10% faster if I somehow eliminated this
|
|
|
|
x = width&~7;
|
|
|
|
filter_line_c(dst+x, src+x, dc+x/2, width-x, thresh, dithers);
|
|
|
|
width = x;
|
|
|
|
}
|
|
|
|
x = -width;
|
2010-03-26 08:13:10 +00:00
|
|
|
__asm__ volatile(
|
2009-07-07 02:01:36 +00:00
|
|
|
"movd %4, %%xmm5 \n"
|
|
|
|
"pxor %%xmm7, %%xmm7 \n"
|
|
|
|
"pshuflw $0,%%xmm5, %%xmm5 \n"
|
|
|
|
"movdqa %6, %%xmm6 \n"
|
|
|
|
"punpcklqdq %%xmm5, %%xmm5 \n"
|
|
|
|
"movdqa %5, %%xmm4 \n"
|
|
|
|
"1: \n"
|
|
|
|
"movq (%2,%0), %%xmm0 \n"
|
|
|
|
"movq (%3,%0), %%xmm1 \n"
|
|
|
|
"punpcklbw %%xmm7, %%xmm0 \n"
|
|
|
|
"punpcklwd %%xmm1, %%xmm1 \n"
|
|
|
|
"psllw $7, %%xmm0 \n"
|
|
|
|
"psubw %%xmm0, %%xmm1 \n" // delta = dc - pix
|
|
|
|
"pabsw %%xmm1, %%xmm2 \n"
|
|
|
|
"pmulhuw %%xmm5, %%xmm2 \n" // m = abs(delta) * thresh >> 16
|
|
|
|
"psubw %%xmm6, %%xmm2 \n"
|
|
|
|
"pminsw %%xmm7, %%xmm2 \n" // m = -max(0, 127-m)
|
|
|
|
"pmullw %%xmm2, %%xmm2 \n"
|
|
|
|
"psllw $1, %%xmm2 \n"
|
|
|
|
"paddw %%xmm4, %%xmm0 \n" // pix += dither
|
|
|
|
"pmulhrsw %%xmm2, %%xmm1 \n" // m = m*m*delta >> 14
|
|
|
|
"paddw %%xmm1, %%xmm0 \n" // pix += m
|
|
|
|
"psraw $7, %%xmm0 \n"
|
|
|
|
"packuswb %%xmm0, %%xmm0 \n"
|
|
|
|
"movq %%xmm0, (%1,%0) \n" // dst = clip(pix>>7)
|
|
|
|
"add $8, %0 \n"
|
|
|
|
"jl 1b \n"
|
|
|
|
:"+&r"(x)
|
|
|
|
:"r"(dst+width), "r"(src+width), "r"(dc+width/2),
|
|
|
|
"rm"(thresh), "m"(*dithers), "m"(*pw_7f)
|
|
|
|
:"memory"
|
|
|
|
);
|
|
|
|
}
|
2010-04-05 16:38:40 +00:00
|
|
|
#endif // HAVE_SSSE3
|
2009-07-07 02:01:36 +00:00
|
|
|
|
2010-04-06 09:46:47 +00:00
|
|
|
#if HAVE_SSE2 && HAVE_6REGS
|
2009-07-07 02:01:36 +00:00
|
|
|
#define BLURV(load)\
|
|
|
|
intptr_t x = -2*width;\
|
2010-03-26 08:13:10 +00:00
|
|
|
__asm__ volatile(\
|
2009-07-07 02:01:36 +00:00
|
|
|
"movdqa %6, %%xmm7 \n"\
|
|
|
|
"1: \n"\
|
|
|
|
load" (%4,%0), %%xmm0 \n"\
|
|
|
|
load" (%5,%0), %%xmm1 \n"\
|
|
|
|
"movdqa %%xmm0, %%xmm2 \n"\
|
|
|
|
"movdqa %%xmm1, %%xmm3 \n"\
|
|
|
|
"psrlw $8, %%xmm0 \n"\
|
|
|
|
"psrlw $8, %%xmm1 \n"\
|
|
|
|
"pand %%xmm7, %%xmm2 \n"\
|
|
|
|
"pand %%xmm7, %%xmm3 \n"\
|
|
|
|
"paddw %%xmm1, %%xmm0 \n"\
|
|
|
|
"paddw %%xmm3, %%xmm2 \n"\
|
|
|
|
"paddw %%xmm2, %%xmm0 \n"\
|
|
|
|
"paddw (%2,%0), %%xmm0 \n"\
|
|
|
|
"movdqa (%1,%0), %%xmm1 \n"\
|
|
|
|
"movdqa %%xmm0, (%1,%0) \n"\
|
|
|
|
"psubw %%xmm1, %%xmm0 \n"\
|
|
|
|
"movdqa %%xmm0, (%3,%0) \n"\
|
|
|
|
"add $16, %0 \n"\
|
|
|
|
"jl 1b \n"\
|
|
|
|
:"+&r"(x)\
|
|
|
|
:"r"(buf+width),\
|
|
|
|
"r"(buf1+width),\
|
|
|
|
"r"(dc+width),\
|
|
|
|
"r"(src+width*2),\
|
|
|
|
"r"(src+width*2+sstride),\
|
|
|
|
"m"(*pw_ff)\
|
|
|
|
:"memory"\
|
|
|
|
);
|
|
|
|
|
|
|
|
static void blur_line_sse2(uint16_t *dc, uint16_t *buf, uint16_t *buf1,
|
|
|
|
uint8_t *src, int sstride, int width)
|
|
|
|
{
|
|
|
|
if (((intptr_t)src|sstride)&15) {
|
|
|
|
BLURV("movdqu");
|
|
|
|
} else {
|
|
|
|
BLURV("movdqa");
|
|
|
|
}
|
|
|
|
}
|
2010-04-05 16:38:40 +00:00
|
|
|
#endif // HAVE_6REGS && HAVE_SSE2
|
2009-07-07 02:01:36 +00:00
|
|
|
|
video/filter: change filter API, use refcounting, remove filter DR
Change the entire filter API to use reference counted images instead
of vf_get_image().
Remove filter "direct rendering". This was useful for vf_expand and (in
rare cases) vf_sub: DR allowed these filters to pass a cropped image to
the filters before them. Then, on filtering, the image was "uncropped",
so that black bars could be added around the image without copying. This
means that in some cases, vf_expand will be slower (-vf gradfun,expand
for example).
Note that another form of DR used for in-place filters has been replaced
by simpler logic. Instead of trying to do DR, filters can check if the
image is writeable (with mp_image_is_writeable()), and do true in-place
if that's the case. This affects filters like vf_gradfun and vf_sub.
Everything has to support strides now. If something doesn't, making a
copy of the image data is required.
2012-11-05 13:25:04 +00:00
|
|
|
static void filter_plane(struct vf_priv_s *ctx, uint8_t *dst, uint8_t *src,
|
|
|
|
int width, int height, int dstride, int sstride, int r)
|
2009-07-07 02:01:36 +00:00
|
|
|
{
|
|
|
|
int bstride = ((width+15)&~15)/2;
|
|
|
|
int y;
|
|
|
|
uint32_t dc_factor = (1<<21)/(r*r);
|
|
|
|
uint16_t *dc = ctx->buf+16;
|
|
|
|
uint16_t *buf = ctx->buf+bstride+32;
|
|
|
|
int thresh = ctx->thresh;
|
|
|
|
|
|
|
|
memset(dc, 0, (bstride+16)*sizeof(*buf));
|
|
|
|
for (y=0; y<r; y++)
|
|
|
|
ctx->blur_line(dc, buf+y*bstride, buf+(y-1)*bstride, src+2*y*sstride, sstride, width/2);
|
|
|
|
for (;;) {
|
|
|
|
if (y < height-r) {
|
|
|
|
int mod = ((y+r)/2)%r;
|
|
|
|
uint16_t *buf0 = buf+mod*bstride;
|
|
|
|
uint16_t *buf1 = buf+(mod?mod-1:r-1)*bstride;
|
|
|
|
int x, v;
|
|
|
|
ctx->blur_line(dc, buf0, buf1, src+(y+r)*sstride, sstride, width/2);
|
|
|
|
for (x=v=0; x<r; x++)
|
|
|
|
v += dc[x];
|
|
|
|
for (; x<width/2; x++) {
|
|
|
|
v += dc[x] - dc[x-r];
|
|
|
|
dc[x-r] = v * dc_factor >> 16;
|
|
|
|
}
|
|
|
|
for (; x<(width+r+1)/2; x++)
|
|
|
|
dc[x-r] = v * dc_factor >> 16;
|
|
|
|
for (x=-r/2; x<0; x++)
|
|
|
|
dc[x] = dc[0];
|
|
|
|
}
|
|
|
|
if (y == r) {
|
|
|
|
for (y=0; y<r; y++)
|
|
|
|
ctx->filter_line(dst+y*dstride, src+y*sstride, dc-r/2, width, thresh, dither[y&7]);
|
|
|
|
}
|
|
|
|
ctx->filter_line(dst+y*dstride, src+y*sstride, dc-r/2, width, thresh, dither[y&7]);
|
|
|
|
if (++y >= height) break;
|
|
|
|
ctx->filter_line(dst+y*dstride, src+y*sstride, dc-r/2, width, thresh, dither[y&7]);
|
|
|
|
if (++y >= height) break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
video/filter: change filter API, use refcounting, remove filter DR
Change the entire filter API to use reference counted images instead
of vf_get_image().
Remove filter "direct rendering". This was useful for vf_expand and (in
rare cases) vf_sub: DR allowed these filters to pass a cropped image to
the filters before them. Then, on filtering, the image was "uncropped",
so that black bars could be added around the image without copying. This
means that in some cases, vf_expand will be slower (-vf gradfun,expand
for example).
Note that another form of DR used for in-place filters has been replaced
by simpler logic. Instead of trying to do DR, filters can check if the
image is writeable (with mp_image_is_writeable()), and do true in-place
if that's the case. This affects filters like vf_gradfun and vf_sub.
Everything has to support strides now. If something doesn't, making a
copy of the image data is required.
2012-11-05 13:25:04 +00:00
|
|
|
static struct mp_image *filter(struct vf_instance *vf, struct mp_image *mpi)
|
2009-07-07 02:01:36 +00:00
|
|
|
{
|
video/filter: change filter API, use refcounting, remove filter DR
Change the entire filter API to use reference counted images instead
of vf_get_image().
Remove filter "direct rendering". This was useful for vf_expand and (in
rare cases) vf_sub: DR allowed these filters to pass a cropped image to
the filters before them. Then, on filtering, the image was "uncropped",
so that black bars could be added around the image without copying. This
means that in some cases, vf_expand will be slower (-vf gradfun,expand
for example).
Note that another form of DR used for in-place filters has been replaced
by simpler logic. Instead of trying to do DR, filters can check if the
image is writeable (with mp_image_is_writeable()), and do true in-place
if that's the case. This affects filters like vf_gradfun and vf_sub.
Everything has to support strides now. If something doesn't, making a
copy of the image data is required.
2012-11-05 13:25:04 +00:00
|
|
|
struct mp_image *dmpi = mpi;
|
|
|
|
if (!mp_image_is_writeable(mpi)) {
|
|
|
|
dmpi = vf_alloc_out_image(vf);
|
|
|
|
mp_image_copy_attributes(dmpi, mpi);
|
2009-07-07 02:01:36 +00:00
|
|
|
}
|
|
|
|
|
video/filter: change filter API, use refcounting, remove filter DR
Change the entire filter API to use reference counted images instead
of vf_get_image().
Remove filter "direct rendering". This was useful for vf_expand and (in
rare cases) vf_sub: DR allowed these filters to pass a cropped image to
the filters before them. Then, on filtering, the image was "uncropped",
so that black bars could be added around the image without copying. This
means that in some cases, vf_expand will be slower (-vf gradfun,expand
for example).
Note that another form of DR used for in-place filters has been replaced
by simpler logic. Instead of trying to do DR, filters can check if the
image is writeable (with mp_image_is_writeable()), and do true in-place
if that's the case. This affects filters like vf_gradfun and vf_sub.
Everything has to support strides now. If something doesn't, making a
copy of the image data is required.
2012-11-05 13:25:04 +00:00
|
|
|
for (int p=0; p < mpi->num_planes; p++) {
|
2009-07-07 02:01:36 +00:00
|
|
|
int w = mpi->w;
|
|
|
|
int h = mpi->h;
|
|
|
|
int r = vf->priv->radius;
|
|
|
|
if (p) {
|
|
|
|
w >>= mpi->chroma_x_shift;
|
|
|
|
h >>= mpi->chroma_y_shift;
|
|
|
|
r = ((r>>mpi->chroma_x_shift) + (r>>mpi->chroma_y_shift)) / 2;
|
|
|
|
r = av_clip((r+1)&~1,4,32);
|
|
|
|
}
|
|
|
|
if (FFMIN(w,h) > 2*r)
|
video/filter: change filter API, use refcounting, remove filter DR
Change the entire filter API to use reference counted images instead
of vf_get_image().
Remove filter "direct rendering". This was useful for vf_expand and (in
rare cases) vf_sub: DR allowed these filters to pass a cropped image to
the filters before them. Then, on filtering, the image was "uncropped",
so that black bars could be added around the image without copying. This
means that in some cases, vf_expand will be slower (-vf gradfun,expand
for example).
Note that another form of DR used for in-place filters has been replaced
by simpler logic. Instead of trying to do DR, filters can check if the
image is writeable (with mp_image_is_writeable()), and do true in-place
if that's the case. This affects filters like vf_gradfun and vf_sub.
Everything has to support strides now. If something doesn't, making a
copy of the image data is required.
2012-11-05 13:25:04 +00:00
|
|
|
filter_plane(vf->priv, dmpi->planes[p], mpi->planes[p], w, h,
|
|
|
|
dmpi->stride[p], mpi->stride[p], r);
|
2009-07-07 02:01:36 +00:00
|
|
|
else if (dmpi->planes[p] != mpi->planes[p])
|
|
|
|
memcpy_pic(dmpi->planes[p], mpi->planes[p], w, h,
|
|
|
|
dmpi->stride[p], mpi->stride[p]);
|
|
|
|
}
|
|
|
|
|
video/filter: change filter API, use refcounting, remove filter DR
Change the entire filter API to use reference counted images instead
of vf_get_image().
Remove filter "direct rendering". This was useful for vf_expand and (in
rare cases) vf_sub: DR allowed these filters to pass a cropped image to
the filters before them. Then, on filtering, the image was "uncropped",
so that black bars could be added around the image without copying. This
means that in some cases, vf_expand will be slower (-vf gradfun,expand
for example).
Note that another form of DR used for in-place filters has been replaced
by simpler logic. Instead of trying to do DR, filters can check if the
image is writeable (with mp_image_is_writeable()), and do true in-place
if that's the case. This affects filters like vf_gradfun and vf_sub.
Everything has to support strides now. If something doesn't, making a
copy of the image data is required.
2012-11-05 13:25:04 +00:00
|
|
|
if (dmpi != mpi)
|
|
|
|
talloc_free(mpi);
|
|
|
|
return dmpi;
|
2009-07-07 02:01:36 +00:00
|
|
|
}
|
|
|
|
|
2009-07-12 12:52:00 +00:00
|
|
|
static int query_format(struct vf_instance *vf, unsigned int fmt)
|
2009-07-07 02:01:36 +00:00
|
|
|
{
|
|
|
|
switch (fmt){
|
|
|
|
case IMGFMT_YVU9:
|
|
|
|
case IMGFMT_IF09:
|
|
|
|
case IMGFMT_YV12:
|
|
|
|
case IMGFMT_I420:
|
|
|
|
case IMGFMT_IYUV:
|
|
|
|
case IMGFMT_CLPL:
|
|
|
|
case IMGFMT_Y800:
|
|
|
|
case IMGFMT_Y8:
|
|
|
|
case IMGFMT_NV12:
|
|
|
|
case IMGFMT_NV21:
|
|
|
|
case IMGFMT_444P:
|
|
|
|
case IMGFMT_422P:
|
|
|
|
case IMGFMT_411P:
|
|
|
|
case IMGFMT_HM12:
|
|
|
|
return vf_next_query_format(vf,fmt);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-07-12 12:52:00 +00:00
|
|
|
static int config(struct vf_instance *vf,
|
2009-07-07 02:01:36 +00:00
|
|
|
int width, int height, int d_width, int d_height,
|
|
|
|
unsigned int flags, unsigned int outfmt)
|
|
|
|
{
|
|
|
|
free(vf->priv->buf);
|
2012-01-08 17:25:26 +00:00
|
|
|
vf->priv->radius = vf->priv->cfg_radius;
|
|
|
|
if (vf->priv->cfg_size > -1) {
|
|
|
|
vf->priv->radius = (vf->priv->cfg_size / 100.0f)
|
|
|
|
* sqrtf(width * width + height * height);
|
|
|
|
}
|
|
|
|
vf->priv->radius = av_clip((vf->priv->radius+1)&~1, 4, 32);
|
2009-07-07 02:01:36 +00:00
|
|
|
vf->priv->buf = av_mallocz((((width+15)&~15)*(vf->priv->radius+1)/2+32)*sizeof(uint16_t));
|
|
|
|
return vf_next_config(vf,width,height,d_width,d_height,flags,outfmt);
|
|
|
|
}
|
|
|
|
|
2009-07-12 12:52:00 +00:00
|
|
|
static void uninit(struct vf_instance *vf)
|
2009-07-07 02:01:36 +00:00
|
|
|
{
|
|
|
|
if (!vf->priv) return;
|
|
|
|
av_free(vf->priv->buf);
|
|
|
|
free(vf->priv);
|
|
|
|
vf->priv = NULL;
|
|
|
|
}
|
|
|
|
|
2010-02-21 13:40:49 +00:00
|
|
|
static int vf_open(vf_instance_t *vf, char *args)
|
2009-07-07 02:01:36 +00:00
|
|
|
{
|
video/filter: change filter API, use refcounting, remove filter DR
Change the entire filter API to use reference counted images instead
of vf_get_image().
Remove filter "direct rendering". This was useful for vf_expand and (in
rare cases) vf_sub: DR allowed these filters to pass a cropped image to
the filters before them. Then, on filtering, the image was "uncropped",
so that black bars could be added around the image without copying. This
means that in some cases, vf_expand will be slower (-vf gradfun,expand
for example).
Note that another form of DR used for in-place filters has been replaced
by simpler logic. Instead of trying to do DR, filters can check if the
image is writeable (with mp_image_is_writeable()), and do true in-place
if that's the case. This affects filters like vf_gradfun and vf_sub.
Everything has to support strides now. If something doesn't, making a
copy of the image data is required.
2012-11-05 13:25:04 +00:00
|
|
|
vf->filter=filter;
|
2009-07-07 02:01:36 +00:00
|
|
|
vf->query_format=query_format;
|
|
|
|
vf->config=config;
|
|
|
|
vf->uninit=uninit;
|
|
|
|
|
2012-01-08 17:25:26 +00:00
|
|
|
bool have_radius = vf->priv->cfg_radius > -1;
|
|
|
|
bool have_size = vf->priv->cfg_size > -1;
|
|
|
|
|
|
|
|
if (have_radius && have_size) {
|
|
|
|
mp_msg(MSGT_VFILTER, MSGL_ERR, "scale: gradfun: only one of "
|
|
|
|
"radius/size parameters allowed at the same time!\n");
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!have_radius && !have_size)
|
2012-01-10 20:18:26 +00:00
|
|
|
vf->priv->cfg_size = 1.0;
|
2012-01-08 17:25:26 +00:00
|
|
|
|
2012-01-08 17:12:43 +00:00
|
|
|
vf->priv->thresh = (1<<15)/av_clipf(vf->priv->cfg_thresh,0.51,255);
|
2009-07-07 02:01:36 +00:00
|
|
|
|
|
|
|
vf->priv->blur_line = blur_line_c;
|
|
|
|
vf->priv->filter_line = filter_line_c;
|
2010-04-06 09:46:47 +00:00
|
|
|
#if HAVE_SSE2 && HAVE_6REGS
|
2009-07-07 02:01:36 +00:00
|
|
|
if (gCpuCaps.hasSSE2)
|
|
|
|
vf->priv->blur_line = blur_line_sse2;
|
|
|
|
#endif
|
2010-04-05 16:38:40 +00:00
|
|
|
#if HAVE_MMX2
|
2009-07-07 02:01:36 +00:00
|
|
|
if (gCpuCaps.hasMMX2)
|
|
|
|
vf->priv->filter_line = filter_line_mmx2;
|
2010-04-05 16:38:40 +00:00
|
|
|
#endif
|
|
|
|
#if HAVE_SSSE3
|
2009-07-07 02:01:36 +00:00
|
|
|
if (gCpuCaps.hasSSSE3)
|
|
|
|
vf->priv->filter_line = filter_line_ssse3;
|
|
|
|
#endif
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2012-01-08 17:12:43 +00:00
|
|
|
#undef ST_OFF
|
|
|
|
#define ST_OFF(f) M_ST_OFF(struct vf_priv_s,f)
|
|
|
|
static const m_option_t vf_opts_fields[] = {
|
|
|
|
{"strength", ST_OFF(cfg_thresh), CONF_TYPE_FLOAT, M_OPT_RANGE, 0.51, 255, NULL},
|
|
|
|
{"radius", ST_OFF(cfg_radius), CONF_TYPE_INT, M_OPT_RANGE, 4, 32, NULL},
|
2012-01-08 17:25:26 +00:00
|
|
|
{"size", ST_OFF(cfg_size), CONF_TYPE_FLOAT, M_OPT_RANGE, 0.1, 5.0, NULL},
|
2012-01-08 17:12:43 +00:00
|
|
|
{ NULL, NULL, 0, 0, 0, 0, NULL }
|
|
|
|
};
|
|
|
|
|
|
|
|
static const m_struct_t vf_opts = {
|
|
|
|
"gradfun",
|
|
|
|
sizeof(struct vf_priv_s),
|
|
|
|
&vf_priv_dflt,
|
|
|
|
vf_opts_fields
|
|
|
|
};
|
|
|
|
|
2009-07-07 02:01:36 +00:00
|
|
|
const vf_info_t vf_info_gradfun = {
|
|
|
|
"gradient deband",
|
|
|
|
"gradfun",
|
|
|
|
"Loren Merritt",
|
|
|
|
"",
|
2010-02-21 13:40:49 +00:00
|
|
|
vf_open,
|
2012-01-08 17:12:43 +00:00
|
|
|
&vf_opts
|
2009-07-07 02:01:36 +00:00
|
|
|
};
|