diff mbox series

[FFmpeg-devel,v13,14/14] avfilter/subscale: Add filter for scaling and/or re-arranging graphical subtitles

Message ID MN2PR04MB598123E0C886782132F557BCBAB59@MN2PR04MB5981.namprd04.prod.outlook.com
State New
Headers show
Series [FFmpeg-devel,v13,01/14] global: Prepare AVFrame for subtitle handling
Related show

Checks

Context Check Description
andriy/make_x86 success Make finished
andriy/make_fate_x86 success Make fate finished
andriy/make_ppc success Make finished
andriy/make_fate_ppc success Make fate finished

Commit Message

Soft Works Oct. 11, 2021, 4:51 a.m. UTC
Signed-off-by: softworkz <softworkz@hotmail.com>
---
 configure                 |   1 +
 doc/filters.texi          | 159 +++++++
 libavfilter/Makefile      |   1 +
 libavfilter/allfilters.c  |   1 +
 libavfilter/sf_subscale.c | 847 ++++++++++++++++++++++++++++++++++++++
 5 files changed, 1009 insertions(+)
 create mode 100644 libavfilter/sf_subscale.c
diff mbox series

Patch

diff --git a/configure b/configure
index 3276264c0c..721a1bc06e 100755
--- a/configure
+++ b/configure
@@ -3669,6 +3669,7 @@  sr_filter_deps="avformat swscale"
 sr_filter_select="dnn"
 stereo3d_filter_deps="gpl"
 splitcc_filter_deps="cc_dec_decoder"
+subscale_filter_deps="swscale avcodec"
 subtitles_filter_deps="avformat avcodec libass"
 super2xsai_filter_deps="gpl"
 pixfmts_super2xsai_test_deps="super2xsai_filter"
diff --git a/doc/filters.texi b/doc/filters.texi
index f18b80ea7c..48c6921a5f 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -25808,6 +25808,165 @@  Set the rendering margin in pixels.
 For rendering, alway use the latest event only, which is covering the given point in time.
 @end table
 
+@section subscale
+
+Provides high-quality scaling and rearranging functionality for graphical subtitles.
+
+The subscale filter provides multiple approaches for manipulating 
+the size and position of graphical subtitle rectangles wich can
+be combined or used separately.
+Scaling is performed by converting the palettized subtitle bitmaps
+to RGBA and re-quantization to palette colors afterwards via elbg algorithm.
+
+The two major operations are 'scale' and 're-arrange' with the 
+latter being separated as 'arrange_h' and 'arrange_v'. 
+
+
+Inputs:
+- 0: Subtitles [bitmap]
+
+Outputs:
+- 0: Subtitles [bitmap]
+
+It accepts the following parameters:
+
+@table @option
+
+@item w, width
+Set the width of the output.
+Width and height in case of graphical subtitles are just indicating 
+a virtual size for which the output (consisting of 0-n bitmap rectangles)
+is intended to be displayed on.
+
+@item h, height
+Set the height of the output.
+
+@item margin_h
+Sets a horizontal margin to be preserverved when using any
+of the arrange modes.
+
+@item margin_v
+Sets a vertical margin to be preserverved when using any
+of the arrange modes.
+
+@item force_original_aspect_ratio
+Enable decreasing or increasing output video width or height if necessary to
+keep the original aspect ratio. Possible values:
+
+@table @samp
+@item disable
+Scale the video as specified and disable this feature.
+
+@item decrease
+The output video dimensions will automatically be decreased if needed.
+
+@item increase
+The output video dimensions will automatically be increased if needed.
+
+@end table
+
+
+@item scale_mode
+Specifies how subtitle bitmaps should be scaled.
+The scale factor is determined by the the factor between input
+and output size.
+
+@table @samp
+@item none
+Do not apply any common scaling.
+
+@item uniform
+Uniformly scale all subtitle bitmaps including their positions.
+
+@item uniform_no_reposition
+Uniformly scale all subtitle bitmaps without changing positions.
+
+@end table
+
+
+@item arrange_h
+Specifies how subtitle bitmaps should be arranged horizontally.
+
+@item arrange_v
+Specifies how subtitle bitmaps should be arranged vertically.
+
+
+@table @samp
+@item none
+Do not rearrange subtitle bitmaps.
+
+@item margin_no_scale
+Move subtitle bitmaps to be positioned inside the specified
+margin (margin_h or margin_v) when possible and without scaling.
+
+@item margin_and_scale
+Move subtitle bitmaps to be positioned inside the specified
+margin (margin_h or margin_v) and scale in case it doesn't fit.
+
+@item snapalign_no_scale
+Categorize subtitle bitmap positions as one of left/center/right
+or top/bottom/middle based on original positioning and apply 
+these alignments for the target positioning. 
+No scaling will be applied.
+
+@item snapalign_and_scale
+Categorize subtitle bitmap positions as one of left/center/right
+or top/bottom/middle based on original positioning and apply 
+these alignments for the target positioning. 
+Bitmaps that do not fit inside the margins borders are 
+scaled to fit.
+@end table
+
+@item eval
+Set evaluation mode for the expressions (@option{width}, @option{height}).
+
+It accepts the following values:
+@table @samp
+@item init
+Evaluate expressions only once during the filter initialization.
+
+@item frame
+Evaluate expressions for each incoming frame. This is way slower than the
+@samp{init} mode since it requires all the scalers to be re-computed, but it
+allows advanced dynamic expressions.
+@end table
+
+Default value is @samp{init}.
+
+
+@item num_colors
+Set the number of palette colors for output images.
+Choose the maximum (256) when further processing is done (e.g. 
+overlaying on a video).
+When subtitles will be encoded as bitmap subtitles (e.g. dvbsub),
+a smaller number of palette colors (e.g. 4-16) might need to be used, depending
+on the target format and codec. 
+
+@item bitmap_width_align
+@item bitmap_height_align
+Make sure that subtitle bitmap sizes are a multiple of this
+value. Default is 2.
+
+@end table
+
+@subsection Examples
+
+@itemize
+@item
+Uniformly scale down video and bitmap subtitles and encode 
+subtitles as dvbsub. 
+@example
+ffmpeg -y -loglevel verbose -ss 32 -i "https://streams.videolan.org/samples/sub/largeres_vobsub.mkv" -filter_complex "[0:0]scale=480x270[vid1];[0:10]subscale=w=480:h=270:num_colors=16[sub1]" -map [vid1] -map [sub1] -c:s dvbsub -c:v libx265 output.ts
+@end example
+@item
+Squeeze video vertically and arrange subtitle bitmaps 
+inside the video area without scaling, then overlay 
+subtitles onto the video.
+@example
+ffmpeg -y -loglevel verbose -ss 32 -i "https://streams.videolan.org/samples/sub/largeres_vobsub.mkv" -filter_complex "[0:0]scale=1920x400,setsar=1[vid1];[0:10]subscale=w=1920:h=400:scale_mode=none:margin_h=50:arrange_h=margin_no_scale:arrange_v=margin_no_scale:margin_v=50:num_colors=256[sub1];[vid1][sub1]overlay_graphicsubs" -c:v libx265 output.ts
+@end example
+@end itemize
+
 @c man end SUBTITLE FILTERS
 
 @chapter Multimedia Filters
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index eb3d2ffca5..4e7c026140 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -549,6 +549,7 @@  OBJS-$(CONFIG_SHOW_SPEAKER_FILTER)           += sf_textmod.o
 OBJS-$(CONFIG_TEXTMOD_FILTER)                += sf_textmod.o
 OBJS-$(CONFIG_SPLITCC_FILTER)                += sf_splitcc.o
 OBJS-$(CONFIG_STRIPSTYLES_FILTER)            += sf_stripstyles.o
+OBJS-$(CONFIG_SUBSCALE_FILTER)               += sf_subscale.o
 
 # multimedia filters
 OBJS-$(CONFIG_ABITSCOPE_FILTER)              += avf_abitscope.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index e94c4f38cb..8fb7d90e65 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -536,6 +536,7 @@  extern const AVFilter ff_sf_graphicsub2text;
 extern const AVFilter ff_sf_showspeaker;
 extern const AVFilter ff_sf_splitcc;
 extern const AVFilter ff_sf_stripstyles;
+extern const AVFilter ff_sf_subscale;
 extern const AVFilter ff_sf_textmod;
 extern const AVFilter ff_svf_graphicsub2video;
 extern const AVFilter ff_svf_textsub2video;
diff --git a/libavfilter/sf_subscale.c b/libavfilter/sf_subscale.c
new file mode 100644
index 0000000000..671a06ddf0
--- /dev/null
+++ b/libavfilter/sf_subscale.c
@@ -0,0 +1,847 @@ 
+/*
+ * Copyright (c) 2021 softworkz
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * scale graphical subtitles filter
+ */
+
+#include <string.h>
+
+#include "drawutils.h"
+#include "internal.h"
+#include "scale_eval.h"
+#include "libavutil/eval.h"
+#include "libavutil/opt.h"
+#include "libswscale/swscale.h"
+
+#include "libavcodec/elbg.h"
+
+static const char *const var_names[] = {
+    "in_w",   "iw",
+    "in_h",   "ih",
+    "out_w",  "ow",
+    "out_h",  "oh",
+    "a",
+    "sar",
+    "dar",
+    "margin_h",
+    "margin_v",
+    NULL
+};
+
+enum var_name {
+    VAR_IN_W,   VAR_IW,
+    VAR_IN_H,   VAR_IH,
+    VAR_OUT_W,  VAR_OW,
+    VAR_OUT_H,  VAR_OH,
+    VAR_A,
+    VAR_SAR,
+    VAR_DAR,
+    VARS_B_H,
+    VARS_B_V,
+    VARS_NB
+};
+
+enum EvalMode {
+    EVAL_MODE_INIT,
+    EVAL_MODE_FRAME,
+    EVAL_MODE_NB
+};
+
+enum SubScaleMode {
+    SSM_NONE,
+    SSM_UNIFORM,
+    SSM_UNIFORM_NO_REPOSITION,
+};
+
+enum SubArrangeMode {
+    SAM_NONE,
+    SAM_ENSUREMARGIN_NO_SCALE,
+    SAM_ENSUREMARGIN_AND_SCALE,
+    SAM_SNAPALIGNMENT_NO_SCALE,
+    SAM_SNAPALIGNMENT_AND_SCALE,
+};
+
+typedef struct SubScaleContext {
+    const AVClass *class;
+    struct SwsContext *sws;
+    AVDictionary *opts;
+
+    int w, h;
+
+    char *w_expr;               ///< width  expression string
+    char *h_expr;               ///< height expression string
+    AVExpr *w_pexpr;
+    AVExpr *h_pexpr;
+    double var_values[VARS_NB];
+
+    int force_original_aspect_ratio;
+    int eval_mode;               ///< expression evaluation mode
+
+    // Scale Options
+    enum SubScaleMode scale_mode;
+
+    // Arrange Options
+    enum SubArrangeMode arrange_mode_h;
+    enum SubArrangeMode arrange_mode_v;
+    int margin_h;
+    int margin_v;
+    char *margin_h_expr; 
+    char *margin_v_expr; 
+    AVExpr *margin_h_pexpr;
+    AVExpr *margin_v_pexpr;
+
+    // Bitmap Options
+    int num_output_colors;
+    int bitmap_width_align;
+    int bitmap_height_align;
+
+    // Color Quantization Fields
+    struct ELBGContext *ctx;
+    AVLFG lfg;
+    int *codeword;
+    int *codeword_closest_codebook_idxs;
+    int *codebook;
+    int r_idx, g_idx, b_idx, a_idx;
+} SubScaleContext;
+
+
+static int config_output(AVFilterLink *outlink);
+
+static int check_exprs(AVFilterContext *ctx)
+{
+    const SubScaleContext *s = ctx->priv;
+    unsigned vars_w[VARS_NB] = { 0 }, vars_h[VARS_NB] = { 0 };
+
+    if (!s->w_pexpr && !s->h_pexpr)
+        return AVERROR(EINVAL);
+
+    if (s->w_pexpr)
+        av_expr_count_vars(s->w_pexpr, vars_w, VARS_NB);
+    if (s->h_pexpr)
+        av_expr_count_vars(s->h_pexpr, vars_h, VARS_NB);
+
+    if (vars_w[VAR_OUT_W] || vars_w[VAR_OW]) {
+        av_log(ctx, AV_LOG_ERROR, "Width expression cannot be self-referencing: '%s'.\n", s->w_expr);
+        return AVERROR(EINVAL);
+    }
+
+    if (vars_h[VAR_OUT_H] || vars_h[VAR_OH]) {
+        av_log(ctx, AV_LOG_ERROR, "Height expression cannot be self-referencing: '%s'.\n", s->h_expr);
+        return AVERROR(EINVAL);
+    }
+
+    if ((vars_w[VAR_OUT_H] || vars_w[VAR_OH]) &&
+        (vars_h[VAR_OUT_W] || vars_h[VAR_OW])) {
+        av_log(ctx, AV_LOG_WARNING, "Circular references detected for width '%s' and height '%s' - possibly invalid.\n", s->w_expr, s->h_expr);
+    }
+
+    if (s->margin_h_pexpr)
+        av_expr_count_vars(s->margin_h_pexpr, vars_w, VARS_NB);
+    if (s->margin_v_pexpr)
+        av_expr_count_vars(s->margin_v_pexpr, vars_h, VARS_NB);
+
+    return 0;
+}
+
+static int scale_parse_expr(AVFilterContext *ctx, char *str_expr, AVExpr **pexpr_ptr, const char *var, const char *args)
+{
+    SubScaleContext *s = ctx->priv;
+    int ret, is_inited = 0;
+    char *old_str_expr = NULL;
+    AVExpr *old_pexpr = NULL;
+
+    if (str_expr) {
+        old_str_expr = av_strdup(str_expr);
+        if (!old_str_expr)
+            return AVERROR(ENOMEM);
+        av_opt_set(s, var, args, 0);
+    }
+
+    if (*pexpr_ptr) {
+        old_pexpr = *pexpr_ptr;
+        *pexpr_ptr = NULL;
+        is_inited = 1;
+    }
+
+    ret = av_expr_parse(pexpr_ptr, args, var_names,
+                        NULL, NULL, NULL, NULL, 0, ctx);
+    if (ret < 0) {
+        av_log(ctx, AV_LOG_ERROR, "Cannot parse expression for %s: '%s'\n", var, args);
+        goto revert;
+    }
+
+    ret = check_exprs(ctx);
+    if (ret < 0)
+        goto revert;
+
+    if (is_inited && (ret = config_output(ctx->outputs[0])) < 0)
+        goto revert;
+
+    av_expr_free(old_pexpr);
+    av_freep(&old_str_expr);
+
+    return 0;
+
+revert:
+    av_expr_free(*pexpr_ptr);
+    *pexpr_ptr = NULL;
+    if (old_str_expr) {
+        av_opt_set(s, var, old_str_expr, 0);
+        av_free(old_str_expr);
+    }
+    if (old_pexpr)
+        *pexpr_ptr = old_pexpr;
+
+    return ret;
+}
+
+static av_cold int init_dict(AVFilterContext *ctx, AVDictionary **opts)
+{
+    SubScaleContext *s = ctx->priv;
+    uint8_t rgba_map[4];
+    int ret;
+
+    if (!s->w_expr)
+        av_opt_set(s, "w", "iw", 0);
+    if (!s->h_expr)
+        av_opt_set(s, "h", "ih", 0);
+
+    ret = scale_parse_expr(ctx, NULL, &s->w_pexpr, "width", s->w_expr);
+    if (ret < 0)
+        return ret;
+
+    ret = scale_parse_expr(ctx, NULL, &s->h_pexpr, "height", s->h_expr);
+    if (ret < 0)
+        return ret;
+
+    av_log(ctx, AV_LOG_VERBOSE, "w:%s h:%s\n",
+           s->w_expr, s->h_expr);
+
+    if (!s->margin_h_expr)
+        av_opt_set(s, "margin_h", "0", 0);
+    if (!s->margin_v_expr)
+        av_opt_set(s, "margin_v", "0", 0);
+
+    ret = scale_parse_expr(ctx, NULL, &s->margin_h_pexpr, "margin_h", s->margin_h_expr);
+    if (ret < 0)
+        return ret;
+
+    ret = scale_parse_expr(ctx, NULL, &s->margin_v_pexpr, "margin_v", s->margin_v_expr);
+    if (ret < 0)
+        return ret;
+
+    s->opts = *opts;
+    *opts = NULL;
+
+    ff_fill_rgba_map(&rgba_map[0], AV_PIX_FMT_RGB32);
+
+    s->r_idx = rgba_map[0]; // R
+    s->g_idx = rgba_map[1]; // G
+    s->b_idx = rgba_map[2]; // B
+    s->a_idx = rgba_map[3]; // A
+
+    av_lfg_init(&s->lfg, 123456789);
+
+
+    return 0;
+}
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+    SubScaleContext *s = ctx->priv;
+    av_expr_free(s->w_pexpr);
+    av_expr_free(s->h_pexpr);
+    s->w_pexpr = s->h_pexpr = NULL;
+
+    av_expr_free(s->margin_h_pexpr);
+    av_expr_free(s->margin_v_pexpr);
+    s->margin_h_pexpr = s->margin_v_pexpr = NULL;
+
+    sws_freeContext(s->sws);
+    s->sws = NULL;
+    av_dict_free(&s->opts);
+
+    avpriv_elbg_free(&s->ctx);
+
+    av_freep(&s->codebook);
+    av_freep(&s->codeword);
+    av_freep(&s->codeword_closest_codebook_idxs);
+}
+
+static int config_input(AVFilterLink *inlink)
+{
+    ////const AVFilterContext *ctx = inlink->dst;
+    ////SubScaleContext *s = ctx->priv;
+
+    ////if (s->w <= 0 || s->h <= 0) {
+    ////    s->w = inlink->w;
+    ////    s->h = inlink->h;
+    ////}
+    return 0;
+}
+
+static int scale_eval_dimensions(AVFilterContext *ctx)
+{
+    SubScaleContext *s = ctx->priv;
+    const AVFilterLink *inlink = ctx->inputs[0];
+    char *expr;
+    int eval_w, eval_h, margin_h, margin_v;
+    int ret;
+    double res;
+
+    s->var_values[VAR_IN_W]  = s->var_values[VAR_IW] = inlink->w;
+    s->var_values[VAR_IN_H]  = s->var_values[VAR_IH] = inlink->h;
+    s->var_values[VAR_OUT_W] = s->var_values[VAR_OW] = NAN;
+    s->var_values[VAR_OUT_H] = s->var_values[VAR_OH] = NAN;
+    s->var_values[VARS_B_H]  = s->var_values[VARS_B_V] = 0;
+    s->var_values[VAR_A]     = (double) inlink->w / inlink->h;
+    s->var_values[VAR_SAR]   = inlink->sample_aspect_ratio.num ?
+        (double) inlink->sample_aspect_ratio.num / inlink->sample_aspect_ratio.den : 1;
+    s->var_values[VAR_DAR]   = s->var_values[VAR_A] * s->var_values[VAR_SAR];
+
+    res = av_expr_eval(s->w_pexpr, s->var_values, NULL);
+    s->var_values[VAR_OUT_W] = s->var_values[VAR_OW] = (int) res == 0 ? inlink->w : (int) res;
+
+    res = av_expr_eval(s->h_pexpr, s->var_values, NULL);
+    if (isnan(res)) {
+        expr = s->h_expr;
+        ret = AVERROR(EINVAL);
+        goto fail;
+    }
+    eval_h = s->var_values[VAR_OUT_H] = s->var_values[VAR_OH] = (int) res == 0 ? inlink->h : (int) res;
+
+    res = av_expr_eval(s->w_pexpr, s->var_values, NULL);
+    if (isnan(res)) {
+        expr = s->w_expr;
+        ret = AVERROR(EINVAL);
+        goto fail;
+    }
+    eval_w = s->var_values[VAR_OUT_W] = s->var_values[VAR_OW] = (int) res == 0 ? inlink->w : (int) res;
+
+    s->w = eval_w;
+    s->h = eval_h;
+
+    res = av_expr_eval(s->margin_h_pexpr, s->var_values, NULL);
+    s->var_values[VARS_B_H] = (int)res;
+
+    res = av_expr_eval(s->margin_v_pexpr, s->var_values, NULL);
+    if (isnan(res)) {
+        expr = s->margin_v_expr;
+        ret = AVERROR(EINVAL);
+        goto fail;
+    }
+    margin_v = s->var_values[VARS_B_V] = (int)res;
+
+    res = av_expr_eval(s->margin_h_pexpr, s->var_values, NULL);
+    if (isnan(res)) {
+        expr = s->margin_h_expr;
+        ret = AVERROR(EINVAL);
+        goto fail;
+    }
+    margin_h = s->var_values[VARS_B_H] = (int)res;
+
+    s->margin_h = margin_h;
+    s->margin_v = margin_v;
+
+    return 0;
+
+fail:
+    av_log(ctx, AV_LOG_ERROR,
+           "Error when evaluating the expression '%s'.\n", expr);
+    return ret;
+}
+
+static int config_output(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    AVFilterLink *inlink  = outlink->src->inputs[0];
+    SubScaleContext *s = ctx->priv;
+    int ret;
+
+    outlink->format = AV_SUBTITLE_FMT_BITMAP;
+    outlink->time_base = ctx->inputs[0]->time_base;
+
+    if ((ret = scale_eval_dimensions(ctx)) < 0)
+        goto fail;
+
+    ff_scale_adjust_dimensions(inlink, &s->w, &s->h,
+                               s->force_original_aspect_ratio, 2);
+
+    if (s->w > INT_MAX ||
+        s->h > INT_MAX ||
+        (s->h * inlink->w) > INT_MAX ||
+        (s->w * inlink->h) > INT_MAX)
+        av_log(ctx, AV_LOG_ERROR, "Rescaled value for width or height is too big.\n");
+
+    outlink->w = s->w;
+    outlink->h = s->h;
+
+    if (s->sws)
+        sws_freeContext(s->sws);
+
+    s->sws = sws_alloc_context();
+    if (!s->sws)
+        return AVERROR(ENOMEM);
+
+    av_opt_set_pixel_fmt(s->sws, "src_format", AV_PIX_FMT_PAL8, 0);
+    av_opt_set_int(s->sws, "dst_format", AV_PIX_FMT_RGB32, 0);
+    av_opt_set_int(s->sws, "threads", ff_filter_get_nb_threads(ctx), 0);
+
+    if (s->opts) {
+        const AVDictionaryEntry *e = NULL;
+        while ((e = av_dict_get(s->opts, "", e, AV_DICT_IGNORE_SUFFIX))) {
+            if ((ret = av_opt_set(s->sws, e->key, e->value, 0)) < 0)
+                return ret;
+        }
+    }
+
+    if ((ret = sws_init_context(s->sws, NULL, NULL)) < 0)
+        return ret;
+
+    if (inlink->sample_aspect_ratio.num){
+        outlink->sample_aspect_ratio = av_mul_q((AVRational){outlink->h * inlink->w, outlink->w * inlink->h}, inlink->sample_aspect_ratio);
+    } else
+        outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
+
+    av_log(ctx, AV_LOG_VERBOSE, "Output size set to %dx%d.\n", outlink->w, outlink->h);
+
+    return 0;
+fail:
+    return ret;
+}
+
+static int palettize_image(SubScaleContext *const s, const int w, const int h, const int src_linesize, uint8_t *src_data,
+                          int dst_linesize, uint8_t *dst_data, uint32_t *dst_pal)
+{
+    int k, ret;
+    const int codeword_length = w * h;
+
+    /* Re-Initialize */
+    s->codeword = av_realloc_f(s->codeword, codeword_length, 4 * sizeof(*s->codeword));
+    if (!s->codeword)
+        return AVERROR(ENOMEM);
+
+    s->codeword_closest_codebook_idxs =
+        av_realloc_f(s->codeword_closest_codebook_idxs, codeword_length,
+                     sizeof(*s->codeword_closest_codebook_idxs));
+    if (!s->codeword_closest_codebook_idxs)
+        return AVERROR(ENOMEM);
+
+    s->codebook = av_realloc_f(s->codebook, s->num_output_colors, 4 * sizeof(*s->codebook));
+    if (!s->codebook)
+        return AVERROR(ENOMEM);
+
+    /* build the codeword */
+    k = 0;
+    for (int i = 0; i < h; i++) {
+        uint8_t *p = src_data;
+        for (int j = 0; j < w; j++) {
+            s->codeword[k++] = p[s->b_idx];
+            s->codeword[k++] = p[s->g_idx];
+            s->codeword[k++] = p[s->r_idx];
+            s->codeword[k++] = p[s->a_idx];
+            p += 4;
+        }
+        src_data += src_linesize;
+    }
+
+    /* compute the codebook */
+    ret = avpriv_elbg_do(&s->ctx, s->codeword, 4,
+        codeword_length, s->codebook,
+        s->num_output_colors, 1,
+        s->codeword_closest_codebook_idxs, &s->lfg, 0);
+
+    if (ret < 0)
+        return ret;
+
+    /* Write Palette */
+    for (int i = 0; i < s->num_output_colors; i++) {
+        dst_pal[i] = s->codebook[i*4+3] << 24  |
+                    (s->codebook[i*4+2] << 16) |
+                    (s->codebook[i*4+1] <<  8) |
+                     s->codebook[i*4  ];
+    }
+
+    /* Write Image */
+    k = 0;
+    for (int i = 0; i < h; i++) {
+        uint8_t *p = dst_data;
+        for (int j = 0; j < w; j++, p++) {
+            p[0] = s->codeword_closest_codebook_idxs[k++];
+        }
+
+        dst_data += dst_linesize;
+    }
+
+    return ret;
+}
+
+static int rescale_size(int64_t a, AVRational factor)
+{
+    const int64_t res = av_rescale_rnd(a, factor.num, factor.den, AV_ROUND_NEAR_INF);
+    if (res > INT32_MAX || res < 0)
+        return 0;
+
+    return (int)res;
+}
+
+
+static int scale_area(AVFilterLink *link, AVSubtitleArea *area, const int target_width, const int target_height)
+{
+    const AVFilterContext *ctx = link->dst;
+    SubScaleContext *s = ctx->priv;
+    int ret;
+
+    AVBufferRef *dst_buffer;
+    const uint8_t* data[2]    = { area->buf[0]->data, (uint8_t *)&area->pal };
+    const int dstW            = FFALIGN(target_width, s->bitmap_width_align);
+    const int dstH            = FFALIGN(target_height, s->bitmap_height_align);
+    const int tmp_linesize[2] = { FFALIGN(dstW * 4, 32), 0 };
+    const int dst_linesize[2] = { dstW, 0 };
+    uint8_t* tmp[2] = { 0, 0 };
+
+    AVBufferRef *tmp_buffer = av_buffer_allocz(tmp_linesize[0] * dstH);
+    if (!tmp_buffer)
+        return AVERROR(ENOMEM);
+
+    if (!s->sws)
+        return 0;
+
+    tmp[0] = tmp_buffer->data;
+
+    s->sws = sws_getCachedContext(s->sws, area->w, area->h, AV_PIX_FMT_PAL8,
+        dstW, dstH, AV_PIX_FMT_RGB32, SWS_BICUBIC, NULL, NULL, NULL);
+    if (!s->sws) {
+        av_log(NULL, AV_LOG_FATAL, "Cannot initialize the conversion context. dstW=%d dstH=%d\n", dstW, dstH);
+        return AVERROR(EINVAL);
+    }
+
+    // Rescale to ARGB
+    ret = sws_scale(s->sws, data, area->linesize, 0, area->h, tmp, tmp_linesize);
+    if (ret < 0) {
+        av_buffer_unref(&tmp_buffer);
+        return ret;
+    }
+
+    // Alloc output buffer
+    dst_buffer = av_buffer_allocz(dst_linesize[0] * dstH);
+    if (!dst_buffer) {
+        av_buffer_unref(&tmp_buffer);
+        return AVERROR(ENOMEM);
+    }
+
+    // Quantize to palettized image
+    ret = palettize_image(s, dstW, dstH, tmp_linesize[0], tmp[0], dst_linesize[0], dst_buffer->data, area->pal);
+    av_buffer_unref(&tmp_buffer);
+
+    if (ret < 0) {
+        av_buffer_unref(&dst_buffer);
+        return ret;
+    }
+
+    av_buffer_unref(&area->buf[0]);
+    ret = av_buffer_replace(&area->buf[0], dst_buffer);
+    if (ret < 0) {
+        av_buffer_unref(&dst_buffer);
+        return ret;
+    }
+
+    area->w = dstW;
+    area->h = dstH;
+    area->linesize[0] = dst_linesize[0];
+    area->nb_colors = s->num_output_colors;
+
+    return ret;
+}
+
+static int process_area(AVFilterLink *inlink, AVSubtitleArea *area, AVRational x_factor, AVRational y_factor)
+{
+    AVFilterContext *ctx     = inlink->dst;
+    const SubScaleContext *s = ctx->priv;
+    int target_w, target_h, target_x, target_y;
+    const int border_l = s->margin_h;
+    const int border_r = s->w - s->margin_h;
+    const int border_t = s->margin_v;
+    const int border_b = s->h - s->margin_v;
+
+    av_log(ctx, AV_LOG_VERBOSE, "process_area -  start: x/y: (%d:%d) size: %dx%d scale_mode: %d x-factor: %d:%d y-factor: %d:%d\n", 
+        area->x, area->y, area->w, area->h, s->scale_mode, x_factor.num, x_factor.den, y_factor.num, y_factor.den);
+
+    switch (s->scale_mode) {
+    case SSM_NONE:
+        target_w = area->w;
+        target_h = area->h;
+        target_x = area->x;
+        target_y = area->y;
+        break;
+    case SSM_UNIFORM:
+        target_w = rescale_size(area->w, x_factor);
+        target_h = rescale_size(area->h, y_factor);
+        target_x = rescale_size(area->x, x_factor);
+        target_y = rescale_size(area->y, y_factor);
+        break;
+    case SSM_UNIFORM_NO_REPOSITION:
+        target_w = rescale_size(area->w, x_factor);
+        target_h = rescale_size(area->h, y_factor);
+        target_x = area->x;
+        target_y = area->y;
+        break;
+    default:
+        av_log(ctx, AV_LOG_ERROR, "Invalid scale_mode: %d\n", s->scale_mode);
+        return AVERROR(EINVAL);
+    }
+
+    av_log(ctx, AV_LOG_VERBOSE, "process_area - scaled: x/y: (%d:%d) size: %dx%d.\n", target_x, target_y, target_w, target_h);
+
+
+    switch (s->arrange_mode_h) {
+    case SAM_ENSUREMARGIN_AND_SCALE:
+    case SAM_SNAPALIGNMENT_AND_SCALE:
+        {
+            // IF it doesn't fit - scale it
+            int max_width = s->w - 2 * s->margin_h;
+
+            if (max_width < 2)
+                max_width = 2;
+
+            if (target_w > max_width) {
+                target_h = (int)av_rescale(target_h, max_width, target_w);
+                target_w = max_width;
+                target_x = s->margin_h;
+            }
+        }
+        break;
+    }
+
+    switch (s->arrange_mode_h) {
+    case SAM_NONE:
+        break;
+    case SAM_ENSUREMARGIN_NO_SCALE:
+    case SAM_ENSUREMARGIN_AND_SCALE:
+        // Left border
+        if (target_x < border_l)
+            target_x = border_l;
+
+        // Right border
+        if (target_x + target_w > border_r)
+            target_x = border_r - target_w;
+
+        break;
+    case SAM_SNAPALIGNMENT_NO_SCALE:
+    case SAM_SNAPALIGNMENT_AND_SCALE:
+        {
+            // Use original values to detect alignment
+            const int left_margin          = area->x;
+            const int right_margin         = inlink->w - area->x - area->w;
+            const AVRational diff_factor_r = { left_margin - right_margin, area->w };
+            const float diff_factor        = (float)av_q2d(diff_factor_r);
+
+            if (diff_factor > 0.2f) {
+                // Right aligned
+                target_x = border_r - target_w;
+            } else if (diff_factor < -0.2f) {
+                // Left aligned
+                target_x = border_l;
+            } else {
+                // Centered
+                target_x = (inlink->w - area->w) / 2;
+            }
+        }
+
+        break;
+    default:
+        av_log(ctx, AV_LOG_ERROR, "Invalid arrange_mode_h: %d\n", s->arrange_mode_h);
+        return AVERROR(EINVAL);
+    }
+
+    av_log(ctx, AV_LOG_VERBOSE, "process_area -  arr_h: x/y: (%d:%d) size: %dx%d.\n", target_x, target_y, target_w, target_h);
+
+
+    switch (s->arrange_mode_v) {
+    case SAM_ENSUREMARGIN_AND_SCALE:
+    case SAM_SNAPALIGNMENT_AND_SCALE:
+        {
+            // IF it doesn't fit - scale it
+            int max_height = s->h - 2 * s->margin_v;
+
+            if (max_height < 2)
+                max_height = 2;
+
+            if (target_h > max_height) {
+                target_w = (int)av_rescale(target_w, max_height, target_h);
+                target_h = max_height;
+                target_y = s->margin_v;
+            }
+        }
+        break;
+    }
+
+    switch (s->arrange_mode_v) {
+    case SAM_NONE:
+        break;
+    case SAM_ENSUREMARGIN_NO_SCALE:
+    case SAM_ENSUREMARGIN_AND_SCALE:
+        // Top border
+        if (target_y < border_t)
+            target_y = border_t;
+
+        // Bottom border
+        if (target_y + target_h > border_b)
+            target_y = border_b - target_h;
+
+        break;
+    case SAM_SNAPALIGNMENT_NO_SCALE:
+    case SAM_SNAPALIGNMENT_AND_SCALE:
+        {
+            // Use original values to detect alignment
+            const int top_margin           = area->y;
+            const int bottom_margin        = inlink->h - area->y - area->h;
+            const AVRational diff_factor_r = { top_margin - bottom_margin, area->h };
+            const float diff_factor        = (float)av_q2d(diff_factor_r);
+
+            if (diff_factor > 0.2f) {
+                // Bottom aligned
+                target_y = border_b - target_h;
+            } else if (diff_factor < -0.2f) {
+                // Top aligned
+                target_y = border_t;
+            } else {
+                // Centered
+                target_y = (inlink->h - area->h) / 2;
+            }
+        }
+
+        break;
+    default:
+        av_log(ctx, AV_LOG_ERROR, "Invalid arrange_mode_v: %d\n", s->arrange_mode_v);
+        return AVERROR(EINVAL);
+    }
+
+    av_log(ctx, AV_LOG_VERBOSE, "process_area -  arr_v: x/y: (%d:%d) size: %dx%d.\n", target_x, target_y, target_w, target_h);
+
+    area->x = target_x;
+    area->y = target_y;
+
+    if (area->w != target_w || area->h != target_h)
+        return scale_area(inlink, area, target_w, target_h);
+
+    return 0;
+}
+
+static int filter_frame(AVFilterLink *inlink, AVFrame *frame)
+{
+    const AVFilterContext *ctx = inlink->dst;
+    AVFilterLink *outlink = ctx->outputs[0];
+    int ret;
+
+    ret = av_frame_make_writable(frame);
+    if (ret >= 0) {
+        const AVRational x_factor = { .num = outlink->w, .den = inlink->w} ;
+        const AVRational y_factor = { .num = outlink->h, .den = inlink->h} ;
+
+        for (unsigned i = 0; i < frame->num_subtitle_areas; ++i) {
+            AVSubtitleArea *area = frame->subtitle_areas[i];
+
+            ret = process_area(inlink, area, x_factor, y_factor);
+            if (ret < 0)
+                return ret;
+        }
+
+        return ff_filter_frame(outlink, frame);
+    }
+
+    return ret;
+}
+
+#define OFFSET(x) offsetof(SubScaleContext, x)
+#define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
+
+static const AVOption subscale_options[] = {
+    { "margin_h", "horizontal border",        OFFSET(margin_h_expr),    AV_OPT_TYPE_STRING,        .flags = FLAGS },
+    { "margin_v", "vertical border",          OFFSET(margin_v_expr),    AV_OPT_TYPE_STRING,        .flags = FLAGS },
+    { "w",     "Output video width",          OFFSET(w_expr),    AV_OPT_TYPE_STRING,        .flags = FLAGS },
+    { "width", "Output video width",          OFFSET(w_expr),    AV_OPT_TYPE_STRING,        .flags = FLAGS },
+    { "h",     "Output video height",         OFFSET(h_expr),    AV_OPT_TYPE_STRING,        .flags = FLAGS },
+    { "height","Output video height",         OFFSET(h_expr),    AV_OPT_TYPE_STRING,        .flags = FLAGS },
+    { "force_original_aspect_ratio", "decrease or increase w/h if necessary to keep the original AR", OFFSET(force_original_aspect_ratio), AV_OPT_TYPE_INT, { .i64 = 0}, 0, 2, FLAGS, "force_oar" },
+    { "disable",  NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 0 }, 0, 0, FLAGS, "force_oar" },
+    { "decrease", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1 }, 0, 0, FLAGS, "force_oar" },
+    { "increase", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2 }, 0, 0, FLAGS, "force_oar" },
+    { "scale_mode", "specify how to scale subtitles", OFFSET(scale_mode), AV_OPT_TYPE_INT, {.i64 = SSM_UNIFORM}, 0, SSM_UNIFORM_NO_REPOSITION, FLAGS, "scale_mode" },
+         { "none",  "no common scaling", 0, AV_OPT_TYPE_CONST, {.i64=SSM_NONE},  .flags = FLAGS, .unit = "scale_mode" },
+         { "uniform",  "uniformly scale and reposition", 0, AV_OPT_TYPE_CONST, {.i64=SSM_UNIFORM},  .flags = FLAGS, .unit = "scale_mode" },
+         { "uniform_no_reposition",  "uniformly scale but keep positions", 0, AV_OPT_TYPE_CONST, {.i64=SSM_UNIFORM_NO_REPOSITION},  .flags = FLAGS, .unit = "scale_mode" },
+
+    { "arrange_h", "specify how to arrange subtitles horizontally", OFFSET(arrange_mode_h), AV_OPT_TYPE_INT, {.i64 = SAM_NONE}, 0, SAM_SNAPALIGNMENT_AND_SCALE, FLAGS, "arrange" },
+    { "arrange_v", "specify how to arrange subtitles vertically", OFFSET(arrange_mode_v), AV_OPT_TYPE_INT, {.i64 = SAM_NONE}, 0, SAM_SNAPALIGNMENT_AND_SCALE, FLAGS, "arrange" },
+         { "none",  "no repositioning", 0, AV_OPT_TYPE_CONST, {.i64=SAM_NONE},  .flags = FLAGS, .unit = "arrange" },
+         { "margin_no_scale",  "move subs inside border when possible", 0, AV_OPT_TYPE_CONST, {.i64=SAM_ENSUREMARGIN_NO_SCALE,},  .flags = FLAGS, .unit = "arrange" },
+         { "margin_and_scale",  "move subs inside border and scale as needed", 0, AV_OPT_TYPE_CONST, {.i64=SAM_ENSUREMARGIN_AND_SCALE,},  .flags = FLAGS, .unit = "arrange" },
+         { "snapalign_no_scale",  "snap subs to near/far/center when possible", 0, AV_OPT_TYPE_CONST, {.i64=SAM_SNAPALIGNMENT_NO_SCALE,},  .flags = FLAGS, .unit = "arrange" },
+         { "snapalign_and_scale", "snap subs to near/far/center and scale as needed", 0, AV_OPT_TYPE_CONST, {.i64=SAM_SNAPALIGNMENT_AND_SCALE,}, .flags = FLAGS, .unit = "arrange" },
+
+    { "eval", "specify when to evaluate expressions", OFFSET(eval_mode), AV_OPT_TYPE_INT, {.i64 = EVAL_MODE_INIT}, 0, EVAL_MODE_NB-1, FLAGS, "eval" },
+         { "init",  "eval expressions once during initialization", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_INIT},  .flags = FLAGS, .unit = "eval" },
+         { "frame", "eval expressions during initialization and per-frame", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_FRAME}, .flags = FLAGS, .unit = "eval" },
+    { "num_colors",     "number of palette colors in output", OFFSET(num_output_colors),  AV_OPT_TYPE_INT, {.i64 = 256 }, 2, 256, FLAGS },
+    { "bitmap_width_align",     "Bitmap width alignment", OFFSET(bitmap_width_align),  AV_OPT_TYPE_INT, {.i64 = 2 }, 1, 256, FLAGS },
+    { "bitmap_height_align",     "Bitmap height alignment", OFFSET(bitmap_height_align),  AV_OPT_TYPE_INT, {.i64 = 2 }, 1, 256, FLAGS },
+    { NULL }
+};
+
+static const AVClass subscale_class = {
+    .class_name       = "subscale",
+    .item_name        = av_default_item_name,
+    .option           = subscale_options,
+    .version          = LIBAVUTIL_VERSION_INT,
+    .category         = AV_CLASS_CATEGORY_FILTER,
+};
+
+static const AVFilterPad inputs[] = {
+    {
+        .name         = "default",
+        .type         = AVMEDIA_TYPE_SUBTITLE,
+        .filter_frame = filter_frame,
+        .config_props = config_input,
+    },
+};
+
+static const AVFilterPad outputs[] = {
+    {
+        .name         = "default",
+        .type         = AVMEDIA_TYPE_SUBTITLE,
+        .config_props = config_output,
+    },
+};
+
+const AVFilter ff_sf_subscale = {
+    .name            = "subscale",
+    .description     = NULL_IF_CONFIG_SMALL("Scale graphical subtitles."),
+    .init_dict       = init_dict,
+    .uninit          = uninit,
+    .priv_size       = sizeof(SubScaleContext),
+    .priv_class      = &subscale_class,
+    FILTER_INPUTS(inputs),
+    FILTER_OUTPUTS(outputs),
+    FILTER_SINGLE_SUBFMT(AV_SUBTITLE_FMT_BITMAP),
+};
+