From: Limin Wang <lance.lmwang at gmail.com>
---
libavfilter/vf_overlay.c | 69 +++++++++++++++++++++++++---------------
1 file changed, 44 insertions(+), 25 deletions(-)
diff --git a/libavfilter/vf_overlay.c b/libavfilter/vf_overlay.c
index ee51a54659..8376494efc 100644
--- a/libavfilter/vf_overlay.c
+++ b/libavfilter/vf_overlay.c
@@ -464,22 +464,26 @@ static av_always_inline void blend_plane_##depth##_##nbits##bits(AVFilterContext
int dst_hp = AV_CEIL_RSHIFT(dst_h, vsub); \
int yp = y>>vsub; \
int xp = x>>hsub; \
- uint8_t *s, *sp, *d, *dp, *dap, *a, *da, *ap; \
+ uint##depth##_t *s, *sp, *d, *dp, *dap, *a, *da, *ap; \
int jmax, j, k, kmax; \
int slice_start, slice_end; \
+ const int max = (1 << nbits) - 1; \
+ const int mid = (1 << (nbits -1)) ; \
+ int bytes = depth / 8; \
\
+ dst_step /= bytes; \
j = FFMAX(-yp, 0); \
jmax = FFMIN3(-yp + dst_hp, FFMIN(src_hp, dst_hp), yp + src_hp); \
\
slice_start = j + (jmax * jobnr) / nb_jobs; \
slice_end = j + (jmax * (jobnr+1)) / nb_jobs; \
\
- sp = src->data[i] + (slice_start) * src->linesize[i]; \
- dp = dst->data[dst_plane] \
+ sp = (uint##depth##_t *)(src->data[i] + (slice_start) * src->linesize[i]); \
+ dp = (uint##depth##_t *)(dst->data[dst_plane] \
+ (yp + slice_start) * dst->linesize[dst_plane] \
- + dst_offset; \
- ap = src->data[3] + (slice_start << vsub) * src->linesize[3]; \
- dap = dst->data[3] + ((yp + slice_start) << vsub) * dst->linesize[3]; \
+ + dst_offset); \
+ ap = (uint##depth##_t *)(src->data[3] + (slice_start << vsub) * src->linesize[3]); \
+ dap = (uint##depth##_t *)(dst->data[3] + ((yp + slice_start) << vsub) * dst->linesize[3]); \
\
for (j = slice_start; j < slice_end; j++) { \
k = FFMAX(-xp, 0); \
@@ -489,7 +493,7 @@ static av_always_inline void blend_plane_##depth##_##nbits##bits(AVFilterContext
da = dap + ((xp+k) << hsub); \
kmax = FFMIN(-xp + dst_wp, src_wp); \
\
- if (((vsub && j+1 < src_hp) || !vsub) && octx->blend_row[i]) { \
+ if (nbits == 8 && ((vsub && j+1 < src_hp) || !vsub) && octx->blend_row[i]) { \
int c = octx->blend_row[i](d, da, s, a, kmax - k, src->linesize[3]); \
\
s += c; \
@@ -515,7 +519,7 @@ static av_always_inline void blend_plane_##depth##_##nbits##bits(AVFilterContext
alpha = a[0]; \
/* if the main channel has an alpha channel, alpha has to be calculated */ \
/* to create an un-premultiplied (straight) alpha value */ \
- if (main_has_alpha && alpha != 0 && alpha != 255) { \
+ if (main_has_alpha && alpha != 0 && alpha != max) { \
/* average alpha for color components, improve quality */ \
uint8_t alpha_d; \
if (hsub && vsub && j+1 < src_hp && k+1 < src_wp) { \
@@ -532,22 +536,32 @@ static av_always_inline void blend_plane_##depth##_##nbits##bits(AVFilterContext
alpha = UNPREMULTIPLY_ALPHA(alpha, alpha_d); \
} \
if (straight) { \
- *d = FAST_DIV255(*d * (255 - alpha) + *s * alpha); \
- } else { \
- if (i && yuv) \
- *d = av_clip(FAST_DIV255((*d - 128) * (255 - alpha)) + *s - 128, -128, 128) + 128; \
+ if (nbits > 8) \
+ *d = (*d * (max - alpha) + *s * alpha) / max; \
else \
- *d = FFMIN(FAST_DIV255(*d * (255 - alpha)) + *s, 255); \
+ *d = FAST_DIV255(*d * (255 - alpha) + *s * alpha); \
+ } else { \
+ if (nbits > 8) { \
+ if (i && yuv) \
+ *d = av_clip((*d * (max - alpha) + *s * alpha) / max + *s - 128, -128, 128) + 128; \
+ else \
+ *d = FFMIN((*d * (max - alpha) + *s * alpha) / max + *s, 255); \
+ } else { \
+ if (i && yuv) \
+ *d = av_clip(FAST_DIV255((*d - 128) * (255 - alpha)) + *s - 128, -128, 128) + 128; \
+ else \
+ *d = FFMIN(FAST_DIV255(*d * (255 - alpha)) + *s, 255); \
+ } \
} \
s++; \
d += dst_step; \
da += 1 << hsub; \
a += 1 << hsub; \
} \
- dp += dst->linesize[dst_plane]; \
- sp += src->linesize[i]; \
- ap += (1 << vsub) * src->linesize[3]; \
- dap += (1 << vsub) * dst->linesize[3]; \
+ dp += dst->linesize[dst_plane] / bytes; \
+ sp += src->linesize[i] / bytes; \
+ ap += (1 << vsub) * src->linesize[3] / bytes; \
+ dap += (1 << vsub) * dst->linesize[3] / bytes; \
} \
}
DEFINE_BLEND_PLANE(8, 8);
@@ -559,18 +573,20 @@ static inline void alpha_composite_##depth##_##nbits##bits(const AVFrame *src, c
int x, int y, \
int jobnr, int nb_jobs) \
{ \
- uint8_t alpha; /* the amount of overlay to blend on to main */ \
- uint8_t *s, *sa, *d, *da; \
+ uint##depth##_t alpha; /* the amount of overlay to blend on to main */ \
+ uint##depth##_t *s, *sa, *d, *da; \
int i, imax, j, jmax; \
int slice_start, slice_end; \
+ const int max = (1 << nbits) - 1; \
+ int bytes = depth / 8; \
\
imax = FFMIN(-y + dst_h, src_h); \
slice_start = (imax * jobnr) / nb_jobs; \
slice_end = ((imax * (jobnr+1)) / nb_jobs); \
\
i = FFMAX(-y, 0); \
- sa = src->data[3] + (i + slice_start) * src->linesize[3]; \
- da = dst->data[3] + (y + i + slice_start) * dst->linesize[3]; \
+ sa = (uint##depth##_t *)(src->data[3] + (i + slice_start) * src->linesize[3]); \
+ da = (uint##depth##_t *)(dst->data[3] + (y + i + slice_start) * dst->linesize[3]); \
\
for (i = i + slice_start; i < slice_end; i++) { \
j = FFMAX(-x, 0); \
@@ -586,18 +602,21 @@ static inline void alpha_composite_##depth##_##nbits##bits(const AVFrame *src, c
switch (alpha) { \
case 0: \
break; \
- case 255: \
+ case max: \
*d = *s; \
break; \
default: \
/* apply alpha compositing: main_alpha += (1-main_alpha) * overlay_alpha */ \
- *d += FAST_DIV255((255 - *d) * *s); \
+ if (nbits > 8) \
+ *d += (max - *d) * *s / max; \
+ else \
+ *d += FAST_DIV255((255 - *d) * *s); \
} \
d += 1; \
s += 1; \
} \
- da += dst->linesize[3]; \
- sa += src->linesize[3]; \
+ da += dst->linesize[3] / bytes; \
+ sa += src->linesize[3] / bytes; \
} \
}
DEFINE_ALPHA_COMPOSITE(8, 8);
--
2.21.0