--- /dev/null
+/*****************************************************************************
+ * gradfun.c: wrapper for the gradfun filter from mplayer
+ *****************************************************************************
+ * Copyright (C) 2010 Laurent Aimar
+ * $Id$
+ *
+ * Authors: Laurent Aimar <fenrir _AT_ videolan _DOT_ org>
+ *
+ * This program 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.
+ *
+ * This program 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 this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
+ *****************************************************************************/
+
+/*****************************************************************************
+ * Preamble
+ *****************************************************************************/
+
+#ifdef HAVE_CONFIG_H
+# include "config.h"
+#endif
+
+#include <vlc_common.h>
+#include <vlc_plugin.h>
+#include <vlc_cpu.h>
+#include <vlc_filter.h>
+
+/*****************************************************************************
+ * Module descriptor
+ *****************************************************************************/
+static int Open (vlc_object_t *);
+static void Close(vlc_object_t *);
+
+#define CFG_PREFIX "gradfun-"
+
+#define RADIUS_MIN (4)
+#define RADIUS_MAX (32)
+#define RADIUS_TEXT N_("Radius")
+#define RADIUS_LONGTEXT N_("Radius in pixels")
+
+#define STRENGTH_MIN (0.51)
+#define STRENGTH_MAX (255)
+#define STRENGTH_TEXT N_("Strength")
+#define STRENGTH_LONGTEXT N_("Strength used to modify the value of a pixel")
+
+vlc_module_begin()
+ set_description(N_("Gradfun video filter"))
+ set_shortname(N_("Gradfun"))
+ set_help("Debanding algorithm")
+ set_capability("video filter2", 0)
+ set_category(CAT_VIDEO)
+ set_subcategory(SUBCAT_VIDEO_VFILTER)
+ add_integer_with_range(CFG_PREFIX "radius", 16, RADIUS_MIN, RADIUS_MAX,
+ NULL, RADIUS_TEXT, RADIUS_LONGTEXT, false)
+ add_float_with_range(CFG_PREFIX "strength", 1.2, STRENGTH_MIN, STRENGTH_MAX,
+ NULL, STRENGTH_TEXT, STRENGTH_LONGTEXT, false)
+
+ set_callbacks(Open, Close)
+vlc_module_end()
+
+/*****************************************************************************
+ * Local prototypes
+ *****************************************************************************/
+#define FFMAX(a,b) __MAX(a,b)
+#ifdef CAN_COMPILE_MMXEXT
+# define HAVE_MMX2 1
+#else
+# define HAVE_MMX2 0
+#endif
+#ifdef CAN_COMPILE_SSE2
+# define HAVE_SSE2 1
+#else
+# define HAVE_SSE2 0
+#endif
+#ifdef CAN_COMPILE_SSSE3
+# define HAVE_SSSE3 1
+#else
+# define HAVE_SSSE3 0
+#endif
+// FIXME too restrictive
+#ifdef __x86_64__
+# define HAVE_6REGS 1
+#else
+# define HAVE_6REGS 0
+#endif
+#define av_clip_uint8 clip_uint8_vlc
+#include "gradfun.h"
+
+static picture_t *Filter(filter_t *, picture_t *);
+static int Callback(vlc_object_t *, char const *, vlc_value_t, vlc_value_t, void *);
+
+struct filter_sys_t {
+ vlc_mutex_t lock;
+ float strength;
+ int radius;
+ int h_shift;
+ int v_shift;
+ void *base_buf;
+ struct vf_priv_s cfg;
+};
+
+static int Open(vlc_object_t *object)
+{
+ filter_t *filter = (filter_t *)object;
+
+ int h_shift;
+ int v_shift;
+ switch (filter->fmt_in.video.i_chroma) {
+ case VLC_CODEC_I410:
+ case VLC_CODEC_YV9:
+ h_shift = 2; v_shift = 2;
+ break;
+ case VLC_CODEC_I411:
+ h_shift = 2; v_shift = 0;
+ break;
+ case VLC_CODEC_I420:
+ case VLC_CODEC_J420:
+ case VLC_CODEC_YV12:
+ h_shift = 1; v_shift = 1;
+ break;
+ case VLC_CODEC_I422:
+ case VLC_CODEC_J422:
+ h_shift = 1; v_shift = 0;
+ break;
+ case VLC_CODEC_I444:
+ case VLC_CODEC_J444:
+ case VLC_CODEC_YUVA:
+ h_shift = 0; v_shift = 0;
+ break;
+ case VLC_CODEC_I440:
+ case VLC_CODEC_J440:
+ h_shift = 0; v_shift = 1;
+ break;
+ default:
+ return VLC_EGENERIC;
+ }
+
+ filter_sys_t *sys = malloc(sizeof(*sys));
+ if (!sys)
+ return VLC_ENOMEM;
+
+ vlc_mutex_init(&sys->lock);
+ sys->h_shift = h_shift;
+ sys->v_shift = v_shift;
+ sys->strength = var_CreateGetFloatCommand(filter, CFG_PREFIX "strength");
+ sys->radius = var_CreateGetIntegerCommand(filter, CFG_PREFIX "radius");
+ var_AddCallback(filter, CFG_PREFIX "strength", Callback, NULL);
+ var_AddCallback(filter, CFG_PREFIX "radius", Callback, NULL);
+ sys->base_buf = NULL;
+
+ struct vf_priv_s *cfg = &sys->cfg;
+ cfg->thresh = 0.0;
+ cfg->radius = 0;
+ cfg->buf = NULL;
+ cfg->filter_line = filter_line_c;
+ cfg->blur_line = blur_line_c;
+
+#if HAVE_SSE2 && HAVE_6REGS
+ if (vlc_CPU() & CPU_CAPABILITY_SSE2)
+ cfg->blur_line = blur_line_sse2;
+#endif
+#if HAVE_MMX2
+ if (vlc_CPU() & CPU_CAPABILITY_MMXEXT)
+ cfg->filter_line = filter_line_mmx2;
+#endif
+#if HAVE_SSSE3
+ if (vlc_CPU() & CPU_CAPABILITY_SSSE3)
+ cfg->filter_line = filter_line_ssse3;
+#endif
+
+ filter->p_sys = sys;
+ filter->pf_video_filter = Filter;
+ return VLC_SUCCESS;
+}
+
+static void Close(vlc_object_t *object)
+{
+ filter_t *filter = (filter_t *)object;
+ filter_sys_t *sys = filter->p_sys;
+
+ free(sys->base_buf);
+ vlc_mutex_destroy(&sys->lock);
+ free(sys);
+}
+
+static picture_t *Filter(filter_t *filter, picture_t *src)
+{
+ filter_sys_t *sys = filter->p_sys;
+
+ picture_t *dst = filter_NewPicture(filter);
+ if (!dst) {
+ picture_Release(src);
+ return NULL;
+ }
+
+ vlc_mutex_lock(&sys->lock);
+ float strength = __MIN(__MAX(sys->strength, STRENGTH_MIN), STRENGTH_MAX);
+ int radius = __MIN(__MAX((sys->radius + 1) & ~1, RADIUS_MIN), RADIUS_MAX);
+ vlc_mutex_unlock(&sys->lock);
+
+ const video_format_t *fmt = &filter->fmt_in.video;
+ struct vf_priv_s *cfg = &sys->cfg;
+
+ cfg->thresh = (1 << 15) / strength;
+ if (cfg->radius != radius) {
+ cfg->radius = radius;
+ cfg->buf = vlc_memalign(&sys->base_buf, 16,
+ (((fmt->i_width + 15) & ~15) * (cfg->radius + 1) / 2 + 32) * sizeof(*cfg->buf));
+ }
+
+ for (int i = 0; i < dst->i_planes; i++) {
+ const plane_t *srcp = &src->p[i];
+ plane_t *dstp = &dst->p[i];
+
+ int w = fmt->i_width;
+ int h = fmt->i_height;
+ int r = cfg->radius;
+ if (i > 0) {
+ w >>= sys->h_shift;
+ h >>= sys->v_shift;
+ r = ((r >> sys->h_shift) + (r >> sys->v_shift)) / 2;
+ r = __MIN(__MAX((r + 1) & ~1, RADIUS_MIN), RADIUS_MAX);
+ }
+ if (__MIN(w, h) > 2 * r && cfg->buf) {
+ filter_plane(cfg, dstp->p_pixels, srcp->p_pixels,
+ w, h, dstp->i_pitch, srcp->i_pitch, r);
+ } else {
+ plane_CopyPixels(dstp, srcp);
+ }
+ }
+
+ picture_CopyProperties(dst, src);
+ picture_Release(src);
+ return dst;
+}
+
+static int Callback(vlc_object_t *object, char const *cmd,
+ vlc_value_t oldval, vlc_value_t newval, void *data)
+{
+ filter_t *filter = (filter_t *)object;
+ filter_sys_t *sys = filter->p_sys;
+ VLC_UNUSED(oldval); VLC_UNUSED(data);
+
+ vlc_mutex_lock(&sys->lock);
+ if (!strcmp(cmd, CFG_PREFIX "strength"))
+ sys->strength = newval.f_float;
+ else
+ sys->radius = newval.i_int;
+ vlc_mutex_unlock(&sys->lock);
+
+ return VLC_SUCCESS;
+}
+
--- /dev/null
+/*
+ * 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.
+ */
+
+struct vf_priv_s {
+ 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);
+};
+
+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;
+ }
+}
+
+#if HAVE_MMX2
+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;
+ __asm__ volatile(
+ "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"
+ );
+}
+#endif
+
+#if HAVE_SSSE3
+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;
+ __asm__ volatile(
+ "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"
+ );
+}
+#endif // HAVE_SSSE3
+
+#if HAVE_SSE2 && HAVE_6REGS
+#define BLURV(load)\
+ intptr_t x = -2*width;\
+ __asm__ volatile(\
+ "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");
+ }
+}
+#endif // HAVE_6REGS && HAVE_SSE2
+
+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)
+{
+ 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;
+ }
+}
+