lavfi: port 2xsai filter from libmpcodecs

Partially based on the port by Niel van der Westhuizen
<nielkie@gmail.com>, done for GCI 2010. Same output as the original
filter and as fast.

See thread:
Subject: [FFmpeg-devel] [PATCH] Port MPlayer 2xSaI filter to libavfilter
Date: Thu, 25 Nov 2010 01:31:24 +1000

Changelog

@@ -25,6 +25,7 @@ version next:
 - drop support for ffmpeg without libavfilter
 - drawtext video filter: fontconfig support
 - ffmpeg -benchmark_all option
+- super2xsai filter ported from libmpcodecs
 
 
 version 0.10:

configure

@@ -1692,6 +1692,7 @@ ocv_filter_deps="libopencv"
 pan_filter_deps="swresample"
 removelogo_filter_deps="avcodec avformat swscale"
 scale_filter_deps="swscale"
+super2xsai_filter_deps="gpl"
 tinterlace_filter_deps="gpl"
 yadif_filter_deps="gpl"
 

doc/filters.texi

@@ -2716,6 +2716,13 @@ For example:
 will create two separate outputs from the same input, one cropped and
 one padded.
 
+@section super2xsai
+
+Scale the input by 2x and smooth using the Super2xSaI (Scale and
+Interpolate) pixel art scaling algorithm.
+
+Useful for enlarging pixel art images without reducing sharpness.
+
 @section swapuv
 Swap U & V plane.
 

libavfilter/Makefile

@@ -94,6 +94,7 @@ OBJS-$(CONFIG_SETTB_FILTER)                  += vf_settb.o
 OBJS-$(CONFIG_SHOWINFO_FILTER)               += vf_showinfo.o
 OBJS-$(CONFIG_SLICIFY_FILTER)                += vf_slicify.o
 OBJS-$(CONFIG_SPLIT_FILTER)                  += vf_split.o
+OBJS-$(CONFIG_SUPER2XSAI_FILTER)             += vf_super2xsai.o
 OBJS-$(CONFIG_SWAPUV_FILTER)                 += vf_swapuv.o
 OBJS-$(CONFIG_THUMBNAIL_FILTER)              += vf_thumbnail.o
 OBJS-$(CONFIG_TILE_FILTER)                   += vf_tile.o

libavfilter/allfilters.c

@@ -99,6 +99,7 @@ void avfilter_register_all(void)
     REGISTER_FILTER (SHOWINFO,    showinfo,    vf);
     REGISTER_FILTER (SLICIFY,     slicify,     vf);
     REGISTER_FILTER (SPLIT,       split,       vf);
+    REGISTER_FILTER (SUPER2XSAI,  super2xsai,  vf);
     REGISTER_FILTER (SWAPUV,      swapuv,      vf);
     REGISTER_FILTER (THUMBNAIL,   thumbnail,   vf);
     REGISTER_FILTER (TILE,        tile,        vf);

libavfilter/version.h

@@ -29,7 +29,7 @@
 #include "libavutil/avutil.h"
 
 #define LIBAVFILTER_VERSION_MAJOR  2
-#define LIBAVFILTER_VERSION_MINOR 68
+#define LIBAVFILTER_VERSION_MINOR 69
 #define LIBAVFILTER_VERSION_MICRO 100
 
 #define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \

libavfilter/vf_super2xsai.c

@@ -0,0 +1,350 @@
+/*
+ * Copyright (c) 2010 Niel van der Westhuizen <nielkie@gmail.com>
+ * Copyright (c) 2002 A'rpi
+ * Copyright (c) 1997-2001 ZSNES Team ( zsknight@zsnes.com / _demo_@zsnes.com )
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * 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 General Public License for more details.
+ *
+ * You should have received a copy of the GNU 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
+ * Super 2xSaI video filter
+ * Ported from MPlayer libmpcodecs/vf_2xsai.c.
+ */
+
+#include "libavutil/pixdesc.h"
+#include "libavutil/intreadwrite.h"
+#include "avfilter.h"
+
+typedef struct {
+    /* masks used for two pixels interpolation */
+    uint32_t hi_pixel_mask;
+    uint32_t lo_pixel_mask;
+
+    /* masks used for four pixels interpolation */
+    uint32_t q_hi_pixel_mask;
+    uint32_t q_lo_pixel_mask;
+
+    int bpp; ///< bytes per pixel, pixel stride for each (packed) pixel
+} Super2xSaIContext;
+
+#define GET_RESULT(A, B, C, D) ((A != C || A != D) - (B != C || B != D))
+
+#define INTERPOLATE(A, B) (((A & hi_pixel_mask) >> 1) + ((B & hi_pixel_mask) >> 1) + (A & B & lo_pixel_mask))
+
+#define Q_INTERPOLATE(A, B, C, D) ((A & q_hi_pixel_mask) >> 2) + ((B & q_hi_pixel_mask) >> 2) + ((C & q_hi_pixel_mask) >> 2) + ((D & q_hi_pixel_mask) >> 2) \
+    + ((((A & q_lo_pixel_mask) + (B & q_lo_pixel_mask) + (C & q_lo_pixel_mask) + (D & q_lo_pixel_mask)) >> 2) & q_lo_pixel_mask)
+
+static void super2xsai(AVFilterContext *ctx,
+                       uint8_t *src, int src_linesize,
+                       uint8_t *dst, int dst_linesize,
+                       int width, int height)
+{
+    Super2xSaIContext *sai = ctx->priv;
+    unsigned int x, y;
+    uint32_t color[4][4];
+    unsigned char *src_line[4];
+    const int bpp = sai->bpp;
+    const uint32_t hi_pixel_mask = sai->hi_pixel_mask;
+    const uint32_t lo_pixel_mask = sai->lo_pixel_mask;
+    const uint32_t q_hi_pixel_mask = sai->q_hi_pixel_mask;
+    const uint32_t q_lo_pixel_mask = sai->q_lo_pixel_mask;
+
+    /* Point to the first 4 lines, first line is duplicated */
+    src_line[0] = src;
+    src_line[1] = src;
+    src_line[2] = src + src_linesize*FFMIN(1, height-1);
+    src_line[3] = src + src_linesize*FFMIN(2, height-1);
+
+#define READ_COLOR4(dst, src_line, off) dst = *((const uint32_t *)src_line + off)
+#define READ_COLOR3(dst, src_line, off) dst = AV_RL24 (src_line + 3*off)
+#define READ_COLOR2(dst, src_line, off) dst = *((const uint16_t *)src_line + off)
+
+    /* Initialise the color matrix for this row. */
+    switch (bpp) {
+    case 4:
+        READ_COLOR4(color[0][0], src_line[0], 0); color[0][1] = color[0][0]; READ_COLOR4(color[0][2], src_line[0], 1); READ_COLOR4(color[0][3], src_line[0], 2);
+        READ_COLOR4(color[1][0], src_line[1], 0); color[1][1] = color[1][0]; READ_COLOR4(color[1][2], src_line[1], 1); READ_COLOR4(color[1][3], src_line[1], 2);
+        READ_COLOR4(color[2][0], src_line[2], 0); color[2][1] = color[2][0]; READ_COLOR4(color[2][2], src_line[2], 1); READ_COLOR4(color[2][3], src_line[2], 2);
+        READ_COLOR4(color[3][0], src_line[3], 0); color[3][1] = color[3][0]; READ_COLOR4(color[3][2], src_line[3], 1); READ_COLOR4(color[3][3], src_line[3], 2);
+        break;
+    case 3:
+        READ_COLOR3(color[0][0], src_line[0], 0); color[0][1] = color[0][0]; READ_COLOR3(color[0][2], src_line[0], 1); READ_COLOR3(color[0][3], src_line[0], 2);
+        READ_COLOR3(color[1][0], src_line[1], 0); color[1][1] = color[1][0]; READ_COLOR3(color[1][2], src_line[1], 1); READ_COLOR3(color[1][3], src_line[1], 2);
+        READ_COLOR3(color[2][0], src_line[2], 0); color[2][1] = color[2][0]; READ_COLOR3(color[2][2], src_line[2], 1); READ_COLOR3(color[2][3], src_line[2], 2);
+        READ_COLOR3(color[3][0], src_line[3], 0); color[3][1] = color[3][0]; READ_COLOR3(color[3][2], src_line[3], 1); READ_COLOR3(color[3][3], src_line[3], 2);
+        break;
+    default:
+        READ_COLOR2(color[0][0], src_line[0], 0); color[0][1] = color[0][0]; READ_COLOR2(color[0][2], src_line[0], 1); READ_COLOR2(color[0][3], src_line[0], 2);
+        READ_COLOR2(color[1][0], src_line[1], 0); color[1][1] = color[1][0]; READ_COLOR2(color[1][2], src_line[1], 1); READ_COLOR2(color[1][3], src_line[1], 2);
+        READ_COLOR2(color[2][0], src_line[2], 0); color[2][1] = color[2][0]; READ_COLOR2(color[2][2], src_line[2], 1); READ_COLOR2(color[2][3], src_line[2], 2);
+        READ_COLOR2(color[3][0], src_line[3], 0); color[3][1] = color[3][0]; READ_COLOR2(color[3][2], src_line[3], 1); READ_COLOR2(color[3][3], src_line[3], 2);
+    }
+
+    for (y = 0; y < height; y++) {
+        uint8_t *dst_line[2];
+
+        dst_line[0] = dst + dst_linesize*2*y;
+        dst_line[1] = dst + dst_linesize*(2*y+1);
+
+        for (x = 0; x < width; x++) {
+            uint32_t product1a, product1b, product2a, product2b;
+
+//---------------------------------------  B0 B1 B2 B3    0  1  2  3
+//                                         4  5* 6  S2 -> 4  5* 6  7
+//                                         1  2  3  S1    8  9 10 11
+//                                         A0 A1 A2 A3   12 13 14 15
+//--------------------------------------
+            if (color[2][1] == color[1][2] && color[1][1] != color[2][2]) {
+                product2b = color[2][1];
+                product1b = product2b;
+            } else if (color[1][1] == color[2][2] && color[2][1] != color[1][2]) {
+                product2b = color[1][1];
+                product1b = product2b;
+            } else if (color[1][1] == color[2][2] && color[2][1] == color[1][2]) {
+                int r = 0;
+
+                r += GET_RESULT(color[1][2], color[1][1], color[1][0], color[3][1]);
+                r += GET_RESULT(color[1][2], color[1][1], color[2][0], color[0][1]);
+                r += GET_RESULT(color[1][2], color[1][1], color[3][2], color[2][3]);
+                r += GET_RESULT(color[1][2], color[1][1], color[0][2], color[1][3]);
+
+                if (r > 0)
+                    product1b = color[1][2];
+                else if (r < 0)
+                    product1b = color[1][1];
+                else
+                    product1b = INTERPOLATE(color[1][1], color[1][2]);
+
+                product2b = product1b;
+            } else {
+                if (color[1][2] == color[2][2] && color[2][2] == color[3][1] && color[2][1] != color[3][2] && color[2][2] != color[3][0])
+                    product2b = Q_INTERPOLATE(color[2][2], color[2][2], color[2][2], color[2][1]);
+                else if (color[1][1] == color[2][1] && color[2][1] == color[3][2] && color[3][1] != color[2][2] && color[2][1] != color[3][3])
+                    product2b = Q_INTERPOLATE(color[2][1], color[2][1], color[2][1], color[2][2]);
+                else
+                    product2b = INTERPOLATE(color[2][1], color[2][2]);
+
+                if (color[1][2] == color[2][2] && color[1][2] == color[0][1] && color[1][1] != color[0][2] && color[1][2] != color[0][0])
+                    product1b = Q_INTERPOLATE(color[1][2], color[1][2], color[1][2], color[1][1]);
+                else if (color[1][1] == color[2][1] && color[1][1] == color[0][2] && color[0][1] != color[1][2] && color[1][1] != color[0][3])
+                    product1b = Q_INTERPOLATE(color[1][2], color[1][1], color[1][1], color[1][1]);
+                else
+                    product1b = INTERPOLATE(color[1][1], color[1][2]);
+            }
+
+            if (color[1][1] == color[2][2] && color[2][1] != color[1][2] && color[1][0] == color[1][1] && color[1][1] != color[3][2])
+                product2a = INTERPOLATE(color[2][1], color[1][1]);
+            else if (color[1][1] == color[2][0] && color[1][2] == color[1][1] && color[1][0] != color[2][1] && color[1][1] != color[3][0])
+                product2a = INTERPOLATE(color[2][1], color[1][1]);
+            else
+                product2a = color[2][1];
+
+            if (color[2][1] == color[1][2] && color[1][1] != color[2][2] && color[2][0] == color[2][1] && color[2][1] != color[0][2])
+                product1a = INTERPOLATE(color[2][1], color[1][1]);
+            else if (color[1][0] == color[2][1] && color[2][2] == color[2][1] && color[2][0] != color[1][1] && color[2][1] != color[0][0])
+                product1a = INTERPOLATE(color[2][1], color[1][1]);
+            else
+                product1a = color[1][1];
+
+            /* Set the calculated pixels */
+            switch (bpp) {
+            case 4:
+                AV_WN32A(dst_line[0] + x * 8,     product1a);
+                AV_WN32A(dst_line[0] + x * 8 + 4, product1b);
+                AV_WN32A(dst_line[1] + x * 8,     product2a);
+                AV_WN32A(dst_line[1] + x * 8 + 4, product2b);
+                break;
+            case 3:
+                AV_WL24(dst_line[0] + x * 6,     product1a);
+                AV_WL24(dst_line[0] + x * 6 + 3, product1b);
+                AV_WL24(dst_line[1] + x * 6,     product2a);
+                AV_WL24(dst_line[1] + x * 6 + 3, product2b);
+                break;
+            default: // bpp = 2
+                AV_WN32A(dst_line[0] + x * 4, product1a | (product1b << 16));
+                AV_WN32A(dst_line[1] + x * 4, product2a | (product2b << 16));
+            }
+
+            /* Move color matrix forward */
+            color[0][0] = color[0][1]; color[0][1] = color[0][2]; color[0][2] = color[0][3];
+            color[1][0] = color[1][1]; color[1][1] = color[1][2]; color[1][2] = color[1][3];
+            color[2][0] = color[2][1]; color[2][1] = color[2][2]; color[2][2] = color[2][3];
+            color[3][0] = color[3][1]; color[3][1] = color[3][2]; color[3][2] = color[3][3];
+
+            if (x < width - 3) {
+                x += 3;
+                switch (bpp) {
+                case 4:
+                    READ_COLOR4(color[0][3], src_line[0], x);
+                    READ_COLOR4(color[1][3], src_line[1], x);
+                    READ_COLOR4(color[2][3], src_line[2], x);
+                    READ_COLOR4(color[3][3], src_line[3], x);
+                    break;
+                case 3:
+                    READ_COLOR3(color[0][3], src_line[0], x);
+                    READ_COLOR3(color[1][3], src_line[1], x);
+                    READ_COLOR3(color[2][3], src_line[2], x);
+                    READ_COLOR3(color[3][3], src_line[3], x);
+                    break;
+                default:        /* case 2 */
+                    READ_COLOR2(color[0][3], src_line[0], x);
+                    READ_COLOR2(color[1][3], src_line[1], x);
+                    READ_COLOR2(color[2][3], src_line[2], x);
+                    READ_COLOR2(color[3][3], src_line[3], x);
+                }
+                x -= 3;
+            }
+        }
+
+        /* We're done with one line, so we shift the source lines up */
+        src_line[0] = src_line[1];
+        src_line[1] = src_line[2];
+        src_line[2] = src_line[3];
+
+        /* Read next line */
+        src_line[3] = src_line[2];
+        if (y < height - 3)
+            src_line[3] += src_linesize;
+
+        switch (bpp) {
+        case 4:
+            READ_COLOR4(color[0][0], src_line[0], 0); color[0][1] = color[0][0]; READ_COLOR4(color[0][2], src_line[0], 1); READ_COLOR4(color[0][3], src_line[0], 2);
+            READ_COLOR4(color[1][0], src_line[1], 0); color[1][1] = color[1][0]; READ_COLOR4(color[1][2], src_line[1], 1); READ_COLOR4(color[1][3], src_line[1], 2);
+            READ_COLOR4(color[2][0], src_line[2], 0); color[2][1] = color[2][1]; READ_COLOR4(color[2][2], src_line[2], 1); READ_COLOR4(color[2][3], src_line[2], 2);
+            READ_COLOR4(color[3][0], src_line[3], 0); color[3][1] = color[3][0]; READ_COLOR4(color[3][2], src_line[3], 1); READ_COLOR4(color[3][3], src_line[3], 2);
+            break;
+        case 3:
+            READ_COLOR3(color[0][0], src_line[0], 0); color[0][1] = color[0][0]; READ_COLOR3(color[0][2], src_line[0], 1); READ_COLOR3(color[0][3], src_line[0], 2);
+            READ_COLOR3(color[1][0], src_line[1], 0); color[1][1] = color[1][0]; READ_COLOR3(color[1][2], src_line[1], 1); READ_COLOR3(color[1][3], src_line[1], 2);
+            READ_COLOR3(color[2][0], src_line[2], 0); color[2][1] = color[2][1]; READ_COLOR3(color[2][2], src_line[2], 1); READ_COLOR3(color[2][3], src_line[2], 2);
+            READ_COLOR3(color[3][0], src_line[3], 0); color[3][1] = color[3][0]; READ_COLOR3(color[3][2], src_line[3], 1); READ_COLOR3(color[3][3], src_line[3], 2);
+            break;
+        default:
+            READ_COLOR2(color[0][0], src_line[0], 0); color[0][1] = color[0][0]; READ_COLOR2(color[0][2], src_line[0], 1); READ_COLOR2(color[0][3], src_line[0], 2);
+            READ_COLOR2(color[1][0], src_line[1], 0); color[1][1] = color[1][0]; READ_COLOR2(color[1][2], src_line[1], 1); READ_COLOR2(color[1][3], src_line[1], 2);
+            READ_COLOR2(color[2][0], src_line[2], 0); color[2][1] = color[2][1]; READ_COLOR2(color[2][2], src_line[2], 1); READ_COLOR2(color[2][3], src_line[2], 2);
+            READ_COLOR2(color[3][0], src_line[3], 0); color[3][1] = color[3][0]; READ_COLOR2(color[3][2], src_line[3], 1); READ_COLOR2(color[3][3], src_line[3], 2);
+        }
+    } // y loop
+}
+
+static int query_formats(AVFilterContext *ctx)
+{
+    static const enum PixelFormat pix_fmts[] = {
+        PIX_FMT_RGBA, PIX_FMT_BGRA, PIX_FMT_ARGB, PIX_FMT_ABGR,
+        PIX_FMT_RGB24, PIX_FMT_BGR24,
+        PIX_FMT_RGB565, PIX_FMT_BGR565, PIX_FMT_RGB555, PIX_FMT_BGR555,
+        PIX_FMT_NONE
+    };
+
+    avfilter_set_common_pixel_formats(ctx, avfilter_make_format_list(pix_fmts));
+    return 0;
+}
+
+static int config_input(AVFilterLink *inlink)
+{
+    Super2xSaIContext *sai = inlink->dst->priv;
+
+    sai->hi_pixel_mask   = 0xFEFEFEFE;
+    sai->lo_pixel_mask   = 0x01010101;
+    sai->q_hi_pixel_mask = 0xFCFCFCFC;
+    sai->q_lo_pixel_mask = 0x03030303;
+    sai->bpp  = 4;
+
+    switch (inlink->format) {
+    case PIX_FMT_RGB24:
+    case PIX_FMT_BGR24:
+        sai->bpp = 3;
+        break;
+
+    case PIX_FMT_RGB565:
+    case PIX_FMT_BGR565:
+        sai->hi_pixel_mask   = 0xF7DEF7DE;
+        sai->lo_pixel_mask   = 0x08210821;
+        sai->q_hi_pixel_mask = 0xE79CE79C;
+        sai->q_lo_pixel_mask = 0x18631863;
+        sai->bpp = 2;
+        break;
+
+    case PIX_FMT_BGR555:
+    case PIX_FMT_RGB555:
+        sai->hi_pixel_mask   = 0x7BDE7BDE;
+        sai->lo_pixel_mask   = 0x04210421;
+        sai->q_hi_pixel_mask = 0x739C739C;
+        sai->q_lo_pixel_mask = 0x0C630C63;
+        sai->bpp = 2;
+        break;
+    }
+
+    return 0;
+}
+
+static int config_output(AVFilterLink *outlink)
+{
+    AVFilterLink *inlink = outlink->src->inputs[0];
+
+    outlink->w = inlink->w*2;
+    outlink->h = inlink->h*2;
+
+    av_log(inlink->dst, AV_LOG_INFO, "fmt:%s size:%dx%d -> size:%dx%d\n",
+           av_get_pix_fmt_name(inlink->format),
+           inlink->w, inlink->h, outlink->w, outlink->h);
+
+    return 0;
+}
+
+static void null_draw_slice(AVFilterLink *inlink, int y, int h, int slice_dir) { }
+
+static void end_frame(AVFilterLink *inlink)
+{
+    AVFilterLink *outlink = inlink->dst->outputs[0];
+    AVFilterBufferRef  *inpicref =  inlink->cur_buf;
+    AVFilterBufferRef *outpicref = outlink->out_buf;
+
+    super2xsai(inlink->dst, inpicref->data[0], inpicref->linesize[0],
+               outpicref->data[0], outpicref->linesize[0],
+               inlink->w, inlink->h);
+
+    avfilter_unref_buffer(inpicref);
+    avfilter_draw_slice(outlink, 0, outlink->h, 1);
+    avfilter_end_frame(outlink);
+    avfilter_unref_buffer(outpicref);
+}
+
+AVFilter avfilter_vf_super2xsai = {
+    .name        = "super2xsai",
+    .description = NULL_IF_CONFIG_SMALL("Scale the input by 2x using the Super2xSaI pixel art algorithm."),
+    .priv_size   = sizeof(Super2xSaIContext),
+    .query_formats = query_formats,
+
+    .inputs = (const AVFilterPad[]) {
+        { .name             = "default",
+          .type             = AVMEDIA_TYPE_VIDEO,
+          .config_props     = config_input,
+          .draw_slice       = null_draw_slice,
+          .end_frame        = end_frame,
+          .min_perms        = AV_PERM_READ },
+        { .name = NULL }
+    },
+    .outputs = (const AVFilterPad[]) {
+        { .name             = "default",
+          .type             = AVMEDIA_TYPE_VIDEO,
+          .config_props     = config_output },
+        { .name = NULL }
+    },
+};