[FFmpeg-devel] [PATCH] lavfi/overlay: add dynamic x/y expression evaluation support
Paul B Mahol
onemda at gmail.com
Wed Feb 20 15:58:51 CET 2013
On 2/19/13, Stefano Sabatini <stefasab at gmail.com> wrote:
> TODO: bump micro
> ---
> doc/filters.texi | 59 ++++++++++++++++++++--------
> libavfilter/vf_overlay.c | 98
> +++++++++++++++++++++++++++++++---------------
> 2 files changed, 109 insertions(+), 48 deletions(-)
>
> diff --git a/doc/filters.texi b/doc/filters.texi
> index 7b45ba4..5dd7404 100644
> --- a/doc/filters.texi
> +++ b/doc/filters.texi
> @@ -3854,33 +3854,53 @@ arguments are interpreted according to the syntax
> @var{x}:@var{y}.
> A description of the accepted options follows.
>
> @table @option
> - at item x, y
> + at item x
> + at item y
> Set the expression for the x and y coordinates of the overlayed video
> -on the main video. Default value is 0.
> +on the main video. These values are evaluated for each new output
> +frame. Default value is "0" for both expressions. In case the
> +expression is invalid, it is set to a huge value (meaning that the
> +overlay will not be displayed within the output visible area).
> +
> + at item rgb
> +If set to 1, force the filter to accept inputs in the RGB
> +color space. Default value is 0.
> +
> + at item shortest
> +If set to 1, force the output to terminate when the shortest input
> +terminates. Default value is 0.
> + at end table
Huh, I saw this in some other patch.
>
> The @var{x} and @var{y} expressions can contain the following
> parameters:
> @table @option
> - at item main_w, main_h
> + at item main_w, W
> + at item main_h, H
> main input width and height
>
> - at item W, H
> -same as @var{main_w} and @var{main_h}
> -
> - at item overlay_w, overlay_h
> + at item overlay_w, w
> + at item overlay_h, h
> overlay input width and height
>
> - at item w, h
> -same as @var{overlay_w} and @var{overlay_h}
> - at end table
> + at item x
> + at item y
> +the computed values for @var{x} and @var{y}. They are evaluated for
> +each new frame.
>
> - at item rgb
> -If set to 1, force the filter to accept inputs in the RGB
> -color space. Default value is 0.
> + at item hsub
> + at item vsub
> +horizontal and vertical chroma subsample values of the output
> +format. For example for the pixel format "yuv422p" @var{hsub} is 2 and
> + at var{vsub} is 1.
>
> - at item shortest
> -If set to 1, force the output to terminate when the shortest input
> -terminates. Default value is 0.
> + at item n
> +the number of input frame, starting from 0
> +
> + at item pos
> +the position in the file of the input frame, NAN if unknown
In what unit? bytes?
> +
> + at item t
> +timestamp expressed in seconds, NAN if the input timestamp is unknown
> @end table
>
> Be aware that frames are taken from each input video in timestamp
> @@ -3941,6 +3961,13 @@ ffplay input.avi -vf 'split[b], pad=iw*2[src],
> [b]deshake, [src]overlay=w'
> @end example
>
> @item
> +Make a sliding overlay appearing from the left to the right top part of
> the
> +screen starting since time 2:
> + at example
> +overlay=x='if(gte(t,2), -w+(t-2)*20, NAN)':y=0
> + at end example
> +
> + at item
> Create a side by side video, combining left and right videos side to side:
> @example
> ffmpeg -i left.avi -i right.avi -filter_complex "
> diff --git a/libavfilter/vf_overlay.c b/libavfilter/vf_overlay.c
> index 607c02d..166c376 100644
> --- a/libavfilter/vf_overlay.c
> +++ b/libavfilter/vf_overlay.c
> @@ -47,6 +47,13 @@ static const char *const var_names[] = {
> "main_h", "H", ///< height of the main video
> "overlay_w", "w", ///< width of the overlay video
> "overlay_h", "h", ///< height of the overlay video
> + "hsub",
> + "vsub",
> + "x",
> + "y",
> + "n", ///< number of frame
> + "pos", ///< position in the file
> + "t", ///< timestamp expressed in seconds
> NULL
> };
>
> @@ -55,6 +62,13 @@ enum var_name {
> VAR_MAIN_H, VAR_MH,
> VAR_OVERLAY_W, VAR_OW,
> VAR_OVERLAY_H, VAR_OH,
> + VAR_HSUB,
> + VAR_VSUB,
> + VAR_X,
> + VAR_Y,
> + VAR_N,
> + VAR_POS,
> + VAR_T,
> VAR_VARS_NB
> };
>
> @@ -93,7 +107,9 @@ typedef struct {
> int hsub, vsub; ///< chroma subsampling values
> int shortest; ///< terminate stream when the shortest
> input terminates
>
> + double var_values[VAR_VARS_NB];
> char *x_expr, *y_expr;
> + AVExpr *x_pexpr, *y_pexpr;
> } OverlayContext;
>
> #define OFFSET(x) offsetof(OverlayContext, x)
> @@ -126,6 +142,9 @@ static av_cold void uninit(AVFilterContext *ctx)
>
> av_opt_free(over);
>
> + av_expr_free(over->x_pexpr); over->x_pexpr = NULL;
> + av_expr_free(over->y_pexpr); over->y_pexpr = NULL;
> +
> avfilter_unref_bufferp(&over->overpicref);
> ff_bufqueue_discard_all(&over->queue_main);
> ff_bufqueue_discard_all(&over->queue_over);
> @@ -194,7 +213,6 @@ static int config_input_overlay(AVFilterLink *inlink)
> AVFilterContext *ctx = inlink->dst;
> OverlayContext *over = inlink->dst->priv;
> char *expr;
> - double var_values[VAR_VARS_NB], res;
> int ret;
> const AVPixFmtDescriptor *pix_desc =
> av_pix_fmt_desc_get(inlink->format);
>
> @@ -202,48 +220,37 @@ static int config_input_overlay(AVFilterLink *inlink)
>
> /* Finish the configuration by evaluating the expressions
> now when both inputs are configured. */
> - var_values[VAR_MAIN_W ] = var_values[VAR_MW] = ctx->inputs[MAIN
> ]->w;
> - var_values[VAR_MAIN_H ] = var_values[VAR_MH] = ctx->inputs[MAIN
> ]->h;
> - var_values[VAR_OVERLAY_W] = var_values[VAR_OW] =
> ctx->inputs[OVERLAY]->w;
> - var_values[VAR_OVERLAY_H] = var_values[VAR_OH] =
> ctx->inputs[OVERLAY]->h;
> -
> - if ((ret = av_expr_parse_and_eval(&res, (expr = over->x_expr),
> var_names, var_values,
> - NULL, NULL, NULL, NULL, NULL, 0,
> ctx)) < 0)
> - goto fail;
> - over->x = res;
> - if ((ret = av_expr_parse_and_eval(&res, (expr = over->y_expr),
> var_names, var_values,
> - NULL, NULL, NULL, NULL, NULL, 0,
> ctx)))
> + over->var_values[VAR_MAIN_W ] = over->var_values[VAR_MW] =
> ctx->inputs[MAIN ]->w;
> + over->var_values[VAR_MAIN_H ] = over->var_values[VAR_MH] =
> ctx->inputs[MAIN ]->h;
> + over->var_values[VAR_OVERLAY_W] = over->var_values[VAR_OW] =
> ctx->inputs[OVERLAY]->w;
> + over->var_values[VAR_OVERLAY_H] = over->var_values[VAR_OH] =
> ctx->inputs[OVERLAY]->h;
> + over->var_values[VAR_HSUB] = 1<<pix_desc->log2_chroma_w;
> + over->var_values[VAR_VSUB] = 1<<pix_desc->log2_chroma_h;
> + over->var_values[VAR_X] = NAN;
> + over->var_values[VAR_Y] = NAN;
> + over->var_values[VAR_N] = 0;
> + over->var_values[VAR_T] = NAN;
> + over->var_values[VAR_POS] = NAN;
> +
> + expr = over->x_expr;
> + if ((ret = av_expr_parse(&over->x_pexpr, expr, var_names,
> + NULL, NULL, NULL, NULL, 0, ctx)) < 0)
> goto fail;
> - over->y = res;
> - /* x may depend on y */
> - if ((ret = av_expr_parse_and_eval(&res, (expr = over->x_expr),
> var_names, var_values,
> - NULL, NULL, NULL, NULL, NULL, 0,
> ctx)) < 0)
> + expr = over->y_expr;
> + if ((ret = av_expr_parse(&over->y_pexpr, expr, var_names,
> + NULL, NULL, NULL, NULL, 0, ctx)) < 0)
> goto fail;
> - over->x = res;
>
> over->overlay_is_packed_rgb =
> ff_fill_rgba_map(over->overlay_rgba_map, inlink->format) >= 0;
> over->overlay_has_alpha = ff_fmt_is_in(inlink->format,
> alpha_pix_fmts);
>
> av_log(ctx, AV_LOG_VERBOSE,
> - "main w:%d h:%d fmt:%s overlay x:%d y:%d w:%d h:%d fmt:%s\n",
> + "main w:%d h:%d fmt:%s overlay w:%d h:%d fmt:%s\n",
> ctx->inputs[MAIN]->w, ctx->inputs[MAIN]->h,
> av_get_pix_fmt_name(ctx->inputs[MAIN]->format),
> - over->x, over->y,
> ctx->inputs[OVERLAY]->w, ctx->inputs[OVERLAY]->h,
> av_get_pix_fmt_name(ctx->inputs[OVERLAY]->format));
> -
> - if (over->x < 0 || over->y < 0 ||
> - over->x + var_values[VAR_OVERLAY_W] > var_values[VAR_MAIN_W] ||
> - over->y + var_values[VAR_OVERLAY_H] > var_values[VAR_MAIN_H]) {
> - av_log(ctx, AV_LOG_WARNING,
> - "Overlay area with coordinates x1:%d y1:%d x2:%d y2:%d "
> - "is not completely contained within the output with size
> %dx%d\n",
> - over->x, over->y,
> - (int)(over->x + var_values[VAR_OVERLAY_W]),
> - (int)(over->y + var_values[VAR_OVERLAY_H]),
> - (int)var_values[VAR_MAIN_W], (int)var_values[VAR_MAIN_H]);
> - }
> return 0;
>
> fail:
> @@ -450,6 +457,7 @@ static int try_filter_frame(AVFilterContext *ctx,
> AVFilterBufferRef *mainpic)
> {
> OverlayContext *over = ctx->priv;
> AVFilterLink *outlink = ctx->outputs[0];
> + AVFilterLink *inlink = ctx->inputs[0];
> AVFilterBufferRef *next_overpic;
> int ret;
>
> @@ -481,8 +489,34 @@ static int try_filter_frame(AVFilterContext *ctx,
> AVFilterBufferRef *mainpic)
> av_ts2str(over->overpicref->pts),
> av_ts2timestr(over->overpicref->pts, &outlink->time_base));
> av_dlog(ctx, "\n");
>
> - if (over->overpicref)
> + if (over->overpicref) {
> + over->var_values[VAR_T] = mainpic->pts == AV_NOPTS_VALUE ?
> + NAN : mainpic->pts * av_q2d(inlink->time_base);
> + over->var_values[VAR_POS] = mainpic->pos == -1 ? NAN :
> mainpic->pos;
> + over->var_values[VAR_X] = av_expr_eval(over->x_pexpr,
> over->var_values, NULL);
> + over->var_values[VAR_Y] = av_expr_eval(over->y_pexpr,
> over->var_values, NULL);
> + over->var_values[VAR_X] = av_expr_eval(over->x_pexpr,
> over->var_values, NULL);
> +
> + if (isnan(over->var_values[VAR_X])) {
> + over->x = INT_MAX;
> + } else {
> + over->x = over->var_values[VAR_X];
> + over->x &= ~((1 << over->hsub) - 1);
> + }
> + if (isnan(over->var_values[VAR_Y])) {
> + over->y = INT_MAX;
> + } else {
> + over->y = over->var_values[VAR_Y];
> + over->y &= ~((1 << over->vsub) - 1);
> + }
> + av_log(ctx, AV_LOG_DEBUG, "n:%f t:%f pos:%f x:%f xi:%d y:%f
> yi:%d\n",
> + over->var_values[VAR_N], over->var_values[VAR_T],
> over->var_values[VAR_POS],
> + over->var_values[VAR_X], over->x,
> + over->var_values[VAR_Y], over->y);
> +
> blend_image(ctx, mainpic, over->overpicref, over->x, over->y);
> + over->var_values[VAR_N] += 1.0;
> + }
> ret = ff_filter_frame(ctx->outputs[0], mainpic);
> av_assert1(ret != AVERROR(EAGAIN));
> over->frame_requested = 0;
> --
> 1.7.9.5
I see nothing obvioulsy wrong, so lgtm.
More information about the ffmpeg-devel
mailing list