avfilter/af_afir: draw IR frequency response

Signed-off-by: Paul B Mahol <onemda@gmail.com>

doc/filters.texi

@@ -969,6 +969,17 @@ Enable applying gain measured from power of IR.
 @item maxir
 Set max allowed Impulse Response filter duration in seconds. Default is 30 seconds.
 Allowed range is 0.1 to 60 seconds.
+
+@item response
+Show IR frequency reponse, magnitude and phase in additional video stream.
+By default it is disabled.
+
+@item channel
+Set for which IR channel to display frequency response. By default is first channel
+displayed. This option is used only when @var{response} is enabled.
+
+@item size
+Set video stream size. This option is used only when @var{response} is enabled.
 @end table
 
 @subsection Examples

libavfilter/af_afir.c

@@ -23,10 +23,14 @@
  * An arbitrary audio FIR filter
  */
 
+#include <float.h>
+
 #include "libavutil/audio_fifo.h"
 #include "libavutil/common.h"
 #include "libavutil/float_dsp.h"
+#include "libavutil/intreadwrite.h"
 #include "libavutil/opt.h"
+#include "libavutil/xga_font_data.h"
 #include "libavcodec/avfft.h"
 
 #include "audio.h"
@@ -162,6 +166,122 @@ static int fir_frame(AudioFIRContext *s, AVFilterLink *outlink)
     return ret;
 }
 
+static void drawtext(AVFrame *pic, int x, int y, const char *txt, uint32_t color)
+{
+    const uint8_t *font;
+    int font_height;
+    int i;
+
+    font = avpriv_cga_font, font_height = 8;
+
+    for (i = 0; txt[i]; i++) {
+        int char_y, mask;
+
+        uint8_t *p = pic->data[0] + y * pic->linesize[0] + (x + i * 8) * 4;
+        for (char_y = 0; char_y < font_height; char_y++) {
+            for (mask = 0x80; mask; mask >>= 1) {
+                if (font[txt[i] * font_height + char_y] & mask)
+                    AV_WL32(p, color);
+                p += 4;
+            }
+            p += pic->linesize[0] - 8 * 4;
+        }
+    }
+}
+
+static void draw_line(AVFrame *out, int x0, int y0, int x1, int y1, uint32_t color)
+{
+    int dx = FFABS(x1-x0), sx = x0 < x1 ? 1 : -1;
+    int dy = FFABS(y1-y0), sy = y0 < y1 ? 1 : -1;
+    int err = (dx>dy ? dx : -dy) / 2, e2;
+
+    for (;;) {
+        AV_WL32(out->data[0] + y0 * out->linesize[0] + x0 * 4, color);
+
+        if (x0 == x1 && y0 == y1)
+            break;
+
+        e2 = err;
+
+        if (e2 >-dx) {
+            err -= dy;
+            x0 += sx;
+        }
+
+        if (e2 < dy) {
+            err += dx;
+            y0 += sy;
+        }
+    }
+}
+
+static void draw_response(AVFilterContext *ctx, AVFrame *out)
+{
+    AudioFIRContext *s = ctx->priv;
+    float *mag, *phase, min = FLT_MAX, max = FLT_MIN;
+    int prev_ymag = -1, prev_yphase = -1;
+    char text[32];
+    int channel, i, x;
+
+    memset(out->data[0], 0, s->h * out->linesize[0]);
+
+    phase = av_malloc_array(s->w, sizeof(*phase));
+    mag = av_malloc_array(s->w, sizeof(*mag));
+    if (!mag || !phase)
+        goto end;
+
+    channel = av_clip(s->ir_channel, 0, s->in[1]->channels);
+    for (i = 0; i < s->w; i++) {
+        const float *src = (const float *)s->in[1]->extended_data[channel];
+        double w = i * M_PI / (s->w - 1);
+        double real = 0.;
+        double imag = 0.;
+
+        for (x = 0; x < s->nb_taps; x++) {
+            real += cos(-x * w) * src[x];
+            imag += sin(-x * w) * src[x];
+        }
+
+        mag[i] = hypot(real, imag);
+        phase[i] = atan2(imag, real);
+        min = fminf(min, mag[i]);
+        max = fmaxf(max, mag[i]);
+    }
+
+    for (i = 0; i < s->w; i++) {
+        int ymag = mag[i] / max * (s->h - 1);
+        int yphase = (0.5 * (1. + phase[i] / M_PI)) * (s->h - 1);
+
+        ymag = s->h - 1 - av_clip(ymag, 0, s->h - 1);
+        yphase = s->h - 1 - av_clip(yphase, 0, s->h - 1);
+
+        if (prev_ymag < 0)
+            prev_ymag = ymag;
+        if (prev_yphase < 0)
+            prev_yphase = yphase;
+
+        draw_line(out, i,   ymag, FFMAX(i - 1, 0),   prev_ymag, 0xFFFF00FF);
+        draw_line(out, i, yphase, FFMAX(i - 1, 0), prev_yphase, 0xFF00FF00);
+
+        prev_ymag   = ymag;
+        prev_yphase = yphase;
+    }
+
+    if (s->w > 400 && s->h > 100) {
+        drawtext(out, 2, 2, "Max Magnitude:", 0xDDDDDDDD);
+        snprintf(text, sizeof(text), "%.2f", max);
+        drawtext(out, 15 * 8 + 2, 2, text, 0xDDDDDDDD);
+
+        drawtext(out, 2, 12, "Min Magnitude:", 0xDDDDDDDD);
+        snprintf(text, sizeof(text), "%.2f", min);
+        drawtext(out, 15 * 8 + 2, 12, text, 0xDDDDDDDD);
+    }
+
+end:
+    av_free(phase);
+    av_free(mag);
+}
+
 static int convert_coeffs(AVFilterContext *ctx)
 {
     AudioFIRContext *s = ctx->priv;
@@ -216,6 +336,9 @@ static int convert_coeffs(AVFilterContext *ctx)
 
     av_audio_fifo_read(s->fifo[1], (void **)s->in[1]->extended_data, s->nb_taps);
 
+    if (s->response)
+        draw_response(ctx, s->video);
+
     if (s->again) {
         float power = 0;
 
@@ -322,6 +445,13 @@ static int filter_frame(AVFilterLink *link, AVFrame *frame)
             return ret;
     }
 
+    if (s->response && s->have_coeffs) {
+        s->video->pts = s->pts;
+        ret = ff_filter_frame(ctx->outputs[1], av_frame_clone(s->video));
+        if (ret < 0)
+            return ret;
+    }
+
     if (s->have_coeffs) {
         while (av_audio_fifo_size(s->fifo[0]) >= s->part_size) {
             ret = fir_frame(s, outlink);
@@ -373,14 +503,26 @@ static int request_frame(AVFilterLink *outlink)
 
 static int query_formats(AVFilterContext *ctx)
 {
+    AudioFIRContext *s = ctx->priv;
     AVFilterFormats *formats;
     AVFilterChannelLayouts *layouts;
     static const enum AVSampleFormat sample_fmts[] = {
         AV_SAMPLE_FMT_FLTP,
         AV_SAMPLE_FMT_NONE
     };
+    static const enum AVPixelFormat pix_fmts[] = {
+        AV_PIX_FMT_RGB0,
+        AV_PIX_FMT_NONE
+    };
     int ret, i;
 
+    if (s->response) {
+        AVFilterLink *videolink = ctx->outputs[1];
+        formats = ff_make_format_list(pix_fmts);
+        if ((ret = ff_formats_ref(formats, &videolink->in_formats)) < 0)
+            return ret;
+    }
+
     layouts = ff_all_channel_counts();
     if ((ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts)) < 0)
         return ret;
@@ -488,11 +630,59 @@ static av_cold void uninit(AVFilterContext *ctx)
     av_audio_fifo_free(s->fifo[1]);
 
     av_freep(&s->fdsp);
+
+    av_freep(&ctx->output_pads[0].name);
+    if (s->response)
+        av_freep(&ctx->output_pads[1].name);
+    av_frame_free(&s->video);
+}
+
+static int config_video(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    AudioFIRContext *s = ctx->priv;
+
+    outlink->sample_aspect_ratio = (AVRational){1,1};
+    outlink->w = s->w;
+    outlink->h = s->h;
+
+    av_frame_free(&s->video);
+    s->video = ff_get_video_buffer(outlink, outlink->w, outlink->h);
+    if (!s->video)
+        return AVERROR(ENOMEM);
+
+    return 0;
 }
 
 static av_cold int init(AVFilterContext *ctx)
 {
     AudioFIRContext *s = ctx->priv;
+    AVFilterPad pad, vpad;
+
+    pad = (AVFilterPad){
+        .name          = av_strdup("default"),
+        .type          = AVMEDIA_TYPE_AUDIO,
+        .config_props  = config_output,
+        .request_frame = request_frame,
+    };
+
+    if (!pad.name)
+        return AVERROR(ENOMEM);
+
+    if (s->response) {
+        vpad = (AVFilterPad){
+            .name         = av_strdup("filter_response"),
+            .type         = AVMEDIA_TYPE_VIDEO,
+            .config_props = config_video,
+        };
+        if (!vpad.name)
+            return AVERROR(ENOMEM);
+    }
+
+    ff_insert_outpad(ctx, 0, &pad);
+
+    if (s->response)
+        ff_insert_outpad(ctx, 1, &vpad);
 
     s->fcmul_add = fcmul_add_c;
 
@@ -519,17 +709,8 @@ static const AVFilterPad afir_inputs[] = {
     { NULL }
 };
 
-static const AVFilterPad afir_outputs[] = {
-    {
-        .name          = "default",
-        .type          = AVMEDIA_TYPE_AUDIO,
-        .config_props  = config_output,
-        .request_frame = request_frame,
-    },
-    { NULL }
-};
-
 #define AF AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
+#define VF AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
 #define OFFSET(x) offsetof(AudioFIRContext, x)
 
 static const AVOption afir_options[] = {
@@ -537,7 +718,10 @@ static const AVOption afir_options[] = {
     { "wet",    "set wet gain",      OFFSET(wet_gain),   AV_OPT_TYPE_FLOAT, {.dbl=1},    0, 10, AF },
     { "length", "set IR length",     OFFSET(length),     AV_OPT_TYPE_FLOAT, {.dbl=1},    0,  1, AF },
     { "again",  "enable auto gain",  OFFSET(again),      AV_OPT_TYPE_BOOL,  {.i64=1},    0,  1, AF },
-    { "maxir",  "set max ir length", OFFSET(max_ir_len), AV_OPT_TYPE_FLOAT, {.dbl=30}, 0.1, 60, AF },
+    { "maxir",  "set max IR length", OFFSET(max_ir_len), AV_OPT_TYPE_FLOAT, {.dbl=30}, 0.1, 60, AF },
+    { "response", "show IR frequency response", OFFSET(response), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, VF },
+    { "channel", "set IR channel to display frequency response", OFFSET(ir_channel), AV_OPT_TYPE_INT, {.i64=0}, 0, 1024, VF },
+    { "size",   "set video size",    OFFSET(w),          AV_OPT_TYPE_IMAGE_SIZE, {.str = "hd720"}, 0, 0, VF },
     { NULL }
 };
 
@@ -552,6 +736,6 @@ AVFilter ff_af_afir = {
     .init          = init,
     .uninit        = uninit,
     .inputs        = afir_inputs,
-    .outputs       = afir_outputs,
-    .flags         = AVFILTER_FLAG_SLICE_THREADS,
+    .flags         = AVFILTER_FLAG_DYNAMIC_OUTPUTS |
+                     AVFILTER_FLAG_SLICE_THREADS,
 };

libavfilter/af_afir.h

@@ -40,6 +40,9 @@ typedef struct AudioFIRContext {
     float length;
     int again;
     float max_ir_len;
+    int response;
+    int w, h;
+    int ir_channel;
 
     float gain;
 
@@ -69,6 +72,7 @@ typedef struct AudioFIRContext {
     AVAudioFifo *fifo[2];
     AVFrame *in[2];
     AVFrame *buffer;
+    AVFrame *video;
     int64_t pts;
     int index;