Canopus HQ/HQA decoder

Based on work by Kostya Shishkov <kostya.shishkov@gmail.com>.

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

1 files changed, 384 insertions, 0 deletions

diff --git a/libavcodec/hq_hqa.c b/libavcodec/hq_hqa.c
new file mode 100644
index 0000000000..b63901aaef
--- /dev/null
+++ b/libavcodec/hq_hqa.c
@@ -0,0 +1,384 @@
+/*
+ * Canopus HQ/HQA decoder
+ *
+ * This file is part of Libav.
+ *
+ * Libav 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.
+ *
+ * Libav 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/intreadwrite.h"
+
+#include "avcodec.h"
+#include "canopus.h"
+#include "internal.h"
+
+#include "hq_hqa.h"
+#include "hq_hqadsp.h"
+
+/* HQ/HQA slices are a set of macroblocks belonging to a frame, and
+ * they usually form a pseudorandom pattern (probably because it is
+ * nicer to display on partial decode).
+ *
+ * For HQA it just happens that each slice is on every 8th macroblock,
+ * but they can be on any frame width like
+ *   X.......X.
+ *   ......X...
+ *   ....X.....
+ *   ..X.......
+ * etc.
+ *
+ * The original decoder has special handling for edge macroblocks,
+ * while lavc simply aligns coded_width and coded_height.
+ */
+
+static inline void put_blocks(HQContext *c, AVFrame *pic,
+                              int plane, int x, int y, int ilace,
+                              int16_t *block0, int16_t *block1)
+{
+    uint8_t *p = pic->data[plane] + x;
+
+    c->hqhqadsp.idct_put(p + y * pic->linesize[plane],
+                         pic->linesize[plane] << ilace, block0);
+    c->hqhqadsp.idct_put(p + (y + (ilace ? 1 : 8)) * pic->linesize[plane],
+                         pic->linesize[plane] << ilace, block1);
+}
+
+static int hq_decode_block(HQContext *c, GetBitContext *gb, int16_t block[64],
+                           int qsel, int is_chroma, int is_hqa)
+{
+    const int32_t *q;
+    int val, pos = 1;
+
+    memset(block, 0, 64 * sizeof(*block));
+
+    if (!is_hqa) {
+        block[0] = get_sbits(gb, 9) << 6;
+        q = ff_hq_quants[qsel][is_chroma][get_bits(gb, 2)];
+    } else {
+        q = ff_hq_quants[qsel][is_chroma][get_bits(gb, 2)];
+        block[0] = get_sbits(gb, 9) << 6;
+    }
+
+    for (;;) {
+        val  = get_vlc2(gb, c->hq_ac_vlc.table, 9, 2);
+        pos += ff_hq_ac_skips[val];
+        if (pos >= 64)
+            break;
+        block[ff_zigzag_direct[pos]] = (ff_hq_ac_syms[val] * q[pos]) >> 12;
+        pos++;
+    }
+
+    return 0;
+}
+
+static int hq_decode_mb(HQContext *c, AVFrame *pic,
+                        GetBitContext *gb, int x, int y)
+{
+    int qgroup, flag;
+    int i, ret;
+
+    qgroup = get_bits(gb, 4);
+    flag = get_bits1(gb);
+
+    for (i = 0; i < 8; i++) {
+        ret = hq_decode_block(c, gb, c->block[i], qgroup, i >= 4, 0);
+        if (ret < 0)
+            return ret;
+    }
+
+    put_blocks(c, pic, 0, x,      y, flag, c->block[0], c->block[2]);
+    put_blocks(c, pic, 0, x + 8,  y, flag, c->block[1], c->block[3]);
+    put_blocks(c, pic, 2, x >> 1, y, flag, c->block[4], c->block[5]);
+    put_blocks(c, pic, 1, x >> 1, y, flag, c->block[6], c->block[7]);
+
+    return 0;
+}
+
+static int hq_decode_frame(HQContext *ctx, AVFrame *pic,
+                           int prof_num, size_t data_size)
+{
+    const HQProfile *profile;
+    GetBitContext gb;
+    const uint8_t *perm, *src = ctx->gbc.buffer;
+    uint32_t slice_off[21];
+    int slice, start_off, next_off, i, ret;
+
+    if (prof_num >= NUM_HQ_PROFILES) {
+        profile = &ff_hq_profile[0];
+        avpriv_request_sample(ctx->avctx, "HQ Profile %d", prof_num);
+    } else {
+        profile = &ff_hq_profile[prof_num];
+        av_log(ctx->avctx, AV_LOG_VERBOSE, "HQ Profile %d\n", prof_num);
+    }
+
+    ctx->avctx->coded_width         = FFALIGN(profile->width,  16);
+    ctx->avctx->coded_height        = FFALIGN(profile->height, 16);
+    ctx->avctx->width               = profile->width;
+    ctx->avctx->height              = profile->height;
+    ctx->avctx->bits_per_raw_sample = 8;
+    ctx->avctx->pix_fmt             = AV_PIX_FMT_YUV422P;
+
+    ret = ff_get_buffer(ctx->avctx, pic, 0);
+    if (ret < 0) {
+        av_log(ctx->avctx, AV_LOG_ERROR, "Could not allocate buffer.\n");
+        return ret;
+    }
+
+    /* Offsets are stored from CUV position, so adjust them accordingly. */
+    for (i = 0; i < profile->num_slices + 1; i++)
+        slice_off[i] = bytestream2_get_be24(&ctx->gbc) - 4;
+
+    next_off = 0;
+    for (slice = 0; slice < profile->num_slices; slice++) {
+        start_off = next_off;
+        next_off  = profile->tab_h * (slice + 1) / profile->num_slices;
+        perm = profile->perm_tab + start_off * profile->tab_w * 2;
+
+        if (slice_off[slice] < (profile->num_slices + 1) * 3 ||
+            slice_off[slice] >= slice_off[slice + 1] ||
+            slice_off[slice + 1] > data_size) {
+            av_log(ctx->avctx, AV_LOG_ERROR,
+                   "Invalid slice size %zu.\n", data_size);
+            break;
+        }
+        init_get_bits(&gb, src + slice_off[slice],
+                      (slice_off[slice + 1] - slice_off[slice]) * 8);
+
+        for (i = 0; i < (next_off - start_off) * profile->tab_w; i++) {
+            ret = hq_decode_mb(ctx, pic, &gb, perm[0] * 16, perm[1] * 16);
+            if (ret < 0) {
+                av_log(ctx->avctx, AV_LOG_ERROR,
+                       "Error decoding macroblock %d at slice %d.\n", i, slice);
+                return ret;
+            }
+            perm += 2;
+        }
+    }
+
+    return 0;
+}
+
+static int hqa_decode_mb(HQContext *c, AVFrame *pic, int qgroup,
+                         GetBitContext *gb, int x, int y)
+{
+    int flag = 0;
+    int i, ret, cbp;
+
+    cbp = get_vlc2(gb, c->hqa_cbp_vlc.table, 5, 1);
+
+    for (i = 0; i < 12; i++)
+        memset(c->block[i], 0, sizeof(*c->block));
+    for (i = 0; i < 12; i++)
+        c->block[i][0] = -128 * (1 << 6);
+
+    if (cbp) {
+        flag = get_bits1(gb);
+
+        cbp |= cbp << 4;
+        if (cbp & 0x3)
+            cbp |= 0x500;
+        if (cbp & 0xC)
+            cbp |= 0xA00;
+        for (i = 0; i < 12; i++) {
+            if (!(cbp & (1 << i)))
+                continue;
+            ret = hq_decode_block(c, gb, c->block[i], qgroup, i >= 8, 1);
+            if (ret < 0)
+                return ret;
+        }
+    }
+
+    put_blocks(c, pic, 3, x,      y, flag, c->block[ 0], c->block[ 2]);
+    put_blocks(c, pic, 3, x + 8,  y, flag, c->block[ 1], c->block[ 3]);
+    put_blocks(c, pic, 0, x,      y, flag, c->block[ 4], c->block[ 6]);
+    put_blocks(c, pic, 0, x + 8,  y, flag, c->block[ 5], c->block[ 7]);
+    put_blocks(c, pic, 2, x >> 1, y, flag, c->block[ 8], c->block[ 9]);
+    put_blocks(c, pic, 1, x >> 1, y, flag, c->block[10], c->block[11]);
+
+    return 0;
+}
+
+static int hqa_decode_slice(HQContext *ctx, AVFrame *pic, GetBitContext *gb,
+                            int quant, int slice_no, int w, int h)
+{
+    int i, j, off;
+    int ret;
+
+    for (i = 0; i < h; i += 16) {
+        off = (slice_no * 16 + i * 3) & 0x70;
+        for (j = off; j < w; j += 128) {
+            ret = hqa_decode_mb(ctx, pic, quant, gb, j, i);
+            if (ret < 0) {
+                av_log(ctx->avctx, AV_LOG_ERROR,
+                       "Error decoding macroblock at %dx%d.\n", i, j);
+                return ret;
+            }
+        }
+    }
+
+    return 0;
+}
+
+static int hqa_decode_frame(HQContext *ctx, AVFrame *pic, size_t data_size)
+{
+    GetBitContext gb;
+    const int num_slices = 8;
+    uint32_t slice_off[9];
+    int i, slice, ret;
+    int width, height, quant;
+    const uint8_t *src = ctx->gbc.buffer;
+
+    width  = bytestream2_get_be16(&ctx->gbc);
+    height = bytestream2_get_be16(&ctx->gbc);
+
+    ctx->avctx->coded_width         = FFALIGN(width,  16);
+    ctx->avctx->coded_height        = FFALIGN(height, 16);
+    ctx->avctx->width               = width;
+    ctx->avctx->height              = height;
+    ctx->avctx->bits_per_raw_sample = 8;
+    ctx->avctx->pix_fmt             = AV_PIX_FMT_YUVA422P;
+
+    av_log(ctx->avctx, AV_LOG_VERBOSE, "HQA Profile\n");
+
+    quant = bytestream2_get_be32(&ctx->gbc);
+    if (quant >= NUM_HQ_QUANTS) {
+        av_log(ctx->avctx, AV_LOG_ERROR,
+               "Invalid quantization matrix %d.\n", quant);
+        return AVERROR_INVALIDDATA;
+    }
+
+    ret = ff_get_buffer(ctx->avctx, pic, 0);
+    if (ret < 0) {
+        av_log(ctx->avctx, AV_LOG_ERROR, "Could not allocate buffer.\n");
+        return ret;
+    }
+
+    /* Offsets are stored from HQA1 position, so adjust them accordingly. */
+    for (i = 0; i < num_slices + 1; i++)
+        slice_off[i] = bytestream2_get_be32(&ctx->gbc) - 4;
+
+    for (slice = 0; slice < num_slices; slice++) {
+        if (slice_off[slice] < (num_slices + 1) * 3 ||
+            slice_off[slice] >= slice_off[slice + 1] ||
+            slice_off[slice + 1] > data_size) {
+            av_log(ctx->avctx, AV_LOG_ERROR,
+                   "Invalid slice size %zu.\n", data_size);
+            break;
+        }
+        init_get_bits(&gb, src + slice_off[slice],
+                      (slice_off[slice + 1] - slice_off[slice]) * 8);
+
+        ret = hqa_decode_slice(ctx, pic, &gb, quant, slice, width, height);
+        if (ret < 0)
+            return ret;
+    }
+
+    return 0;
+}
+
+static int hq_hqa_decode_frame(AVCodecContext *avctx, void *data,
+                               int *got_frame, AVPacket *avpkt)
+{
+    HQContext *ctx = avctx->priv_data;
+    AVFrame *pic = data;
+    uint32_t info_tag;
+    unsigned int data_size;
+    int tag, ret;
+
+    bytestream2_init(&ctx->gbc, avpkt->data, avpkt->size);
+    if (bytestream2_get_bytes_left(&ctx->gbc) < 4 + 4) {
+        av_log(avctx, AV_LOG_ERROR, "Frame is too small (%d).\n", avpkt->size);
+        return AVERROR_INVALIDDATA;
+    }
+
+    info_tag = bytestream2_get_le32(&ctx->gbc);
+    if (info_tag == MKTAG('I', 'N', 'F', 'O')) {
+        int info_size = bytestream2_get_le32(&ctx->gbc);
+        if (bytestream2_get_bytes_left(&ctx->gbc) < info_size) {
+            av_log(avctx, AV_LOG_ERROR, "Invalid INFO size (%d).\n", info_size);
+            return AVERROR_INVALIDDATA;
+        }
+        ff_canopus_parse_info_tag(avctx, ctx->gbc.buffer, info_size);
+
+        bytestream2_skip(&ctx->gbc, info_size);
+    }
+
+    data_size = bytestream2_get_bytes_left(&ctx->gbc);
+    if (data_size < 4) {
+        av_log(avctx, AV_LOG_ERROR, "Frame is too small (%d).\n", data_size);
+        return AVERROR_INVALIDDATA;
+    }
+
+    /* HQ defines dimensions and number of slices, and thus slice traversal
+     * order. HQA has no size constraint and a fixed number of slices, so it
+     * needs a separate scheme for it. */
+    tag = bytestream2_get_le32(&ctx->gbc);
+    if ((tag & 0x00FFFFFF) == (MKTAG('U', 'V', 'C', ' ') & 0x00FFFFFF)) {
+        ret = hq_decode_frame(ctx, pic, tag >> 24, data_size);
+    } else if (tag == MKTAG('H', 'Q', 'A', '1')) {
+        ret = hqa_decode_frame(ctx, pic, data_size);
+    } else {
+        av_log(avctx, AV_LOG_ERROR, "Not a HQ/HQA frame.\n");
+        return AVERROR_INVALIDDATA;
+    }
+    if (ret < 0) {
+        av_log(avctx, AV_LOG_ERROR, "Error decoding frame.\n");
+        return ret;
+    }
+
+    pic->key_frame = 1;
+    pic->pict_type = AV_PICTURE_TYPE_I;
+
+    *got_frame = 1;
+
+    return avpkt->size;
+}
+
+static av_cold int hq_hqa_decode_init(AVCodecContext *avctx)
+{
+    HQContext *ctx = avctx->priv_data;
+    ctx->avctx = avctx;
+
+    ff_hqdsp_init(&ctx->hqhqadsp);
+
+    return ff_hq_init_vlcs(ctx);
+}
+
+static av_cold int hq_hqa_decode_close(AVCodecContext *avctx)
+{
+    HQContext *ctx = avctx->priv_data;
+
+    ff_free_vlc(&ctx->hq_ac_vlc);
+    ff_free_vlc(&ctx->hqa_cbp_vlc);
+
+    return 0;
+}
+
+AVCodec ff_hq_hqa_decoder = {
+    .name           = "hq_hqa",
+    .long_name      = NULL_IF_CONFIG_SMALL("Canopus HQ/HQA"),
+    .type           = AVMEDIA_TYPE_VIDEO,
+    .id             = AV_CODEC_ID_HQ_HQA,
+    .priv_data_size = sizeof(HQContext),
+    .init           = hq_hqa_decode_init,
+    .decode         = hq_hqa_decode_frame,
+    .close          = hq_hqa_decode_close,
+    .capabilities   = CODEC_CAP_DR1,
+    .caps_internal  = FF_CODEC_CAP_INIT_THREADSAFE |
+                      FF_CODEC_CAP_INIT_CLEANUP,
+};