annotate rtjpeg.c @ 7351:1502ba3beb72 libavcodec

The codebook generator algorithm involves picking three different codebook centroids ("high utility", "low utility" and "closest to the low utility one"). This change avoid the corner case of choosing two times the same centroid.

/*
 * RTJpeg decoding functions
 * Copyright (c) 2006 Reimar Doeffinger
 *
 * 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
 */
#include "libavutil/common.h"
#include "bitstream.h"
#include "dsputil.h"
#include "rtjpeg.h"

#define PUT_COEFF(c) \
    i = scan[coeff--]; \
    block[i] = (c) * quant[i];

//! aligns the bitstream to the give power of two
#define ALIGN(a) \
    n = (-get_bits_count(gb)) & (a - 1); \
    if (n) {skip_bits(gb, n);}

/**
 * \brief read one block from stream
 * \param gb contains stream data
 * \param block where data is written to
 * \param scan array containing the mapping stream address -> block position
 * \param quant quantization factors
 *
 * Note: GetBitContext is used to make the code simpler, since all data is
 * aligned this could be done faster in a different way, e.g. as it is done
 * in MPlayer libmpcodecs/native/rtjpegn.c.
 */
static inline int get_block(GetBitContext *gb, DCTELEM *block, const uint8_t *scan,
                            const uint32_t *quant) {
    int coeff, i, n;
    int8_t ac;
    uint8_t dc = get_bits(gb, 8);

    // block not coded
    if (dc == 255)
       return 0;

    // number of non-zero coefficients
    coeff = get_bits(gb, 6);
    // normally we would only need to clear the (63 - coeff) last values,
    // but since we do not know where they are we just clear the whole block
    memset(block, 0, 64 * sizeof(DCTELEM));

    // 2 bits per coefficient
    while (coeff) {
        ac = get_sbits(gb, 2);
        if (ac == -2)
            break; // continue with more bits
        PUT_COEFF(ac);
    }

    // 4 bits per coefficient
    ALIGN(4);
    while (coeff) {
        ac = get_sbits(gb, 4);
        if (ac == -8)
            break; // continue with more bits
        PUT_COEFF(ac);
    }

    // 8 bits per coefficient
    ALIGN(8);
    while (coeff) {
        ac = get_sbits(gb, 8);
        PUT_COEFF(ac);
    }

    PUT_COEFF(dc);
    return 1;
}

/**
 * \brief decode one rtjpeg YUV420 frame
 * \param c context, must be initialized via rtjpeg_decode_init
 * \param f AVFrame to place decoded frame into. If parts of the frame
 *          are not coded they are left unchanged, so consider initializing it
 * \param buf buffer containing input data
 * \param buf_size length of input data in bytes
 * \return number of bytes consumed from the input buffer
 */
int rtjpeg_decode_frame_yuv420(RTJpegContext *c, AVFrame *f,
                               const uint8_t *buf, int buf_size) {
    DECLARE_ALIGNED_16(DCTELEM, block[64]);
    GetBitContext gb;
    int w = c->w / 16, h = c->h / 16;
    int x, y;
    uint8_t *y1 = f->data[0], *y2 = f->data[0] + 8 * f->linesize[0];
    uint8_t *u = f->data[1], *v = f->data[2];
    init_get_bits(&gb, buf, buf_size * 8);
    for (y = 0; y < h; y++) {
        for (x = 0; x < w; x++) {
            if (get_block(&gb, block, c->scan, c->lquant))
                c->dsp->idct_put(y1, f->linesize[0], block);
            y1 += 8;
            if (get_block(&gb, block, c->scan, c->lquant))
                c->dsp->idct_put(y1, f->linesize[0], block);
            y1 += 8;
            if (get_block(&gb, block, c->scan, c->lquant))
                c->dsp->idct_put(y2, f->linesize[0], block);
            y2 += 8;
            if (get_block(&gb, block, c->scan, c->lquant))
                c->dsp->idct_put(y2, f->linesize[0], block);
            y2 += 8;
            if (get_block(&gb, block, c->scan, c->cquant))
                c->dsp->idct_put(u, f->linesize[1], block);
            u += 8;
            if (get_block(&gb, block, c->scan, c->cquant))
                c->dsp->idct_put(v, f->linesize[2], block);
            v += 8;
        }
        y1 += 2 * 8 * (f->linesize[0] - w);
        y2 += 2 * 8 * (f->linesize[0] - w);
        u += 8 * (f->linesize[1] - w);
        v += 8 * (f->linesize[2] - w);
    }
    return get_bits_count(&gb) / 8;
}

/**
 * \brief initialize an RTJpegContext, may be called multiple times
 * \param c context to initialize
 * \param dsp specifies the idct to use for decoding
 * \param width width of image, will be rounded down to the nearest multiple
 *              of 16 for decoding
 * \param height height of image, will be rounded down to the nearest multiple
 *              of 16 for decoding
 * \param lquant luma quantization table to use
 * \param cquant chroma quantization table to use
 */
void rtjpeg_decode_init(RTJpegContext *c, DSPContext *dsp,
                        int width, int height,
                        const uint32_t *lquant, const uint32_t *cquant) {
    int i;
    c->dsp = dsp;
    for (i = 0; i < 64; i++) {
        int z = ff_zigzag_direct[i];
        int p = c->dsp->idct_permutation[i];
        z = ((z << 3) | (z >> 3)) & 63; // rtjpeg uses a transposed variant

        // permute the scan and quantization tables for the chosen idct
        c->scan[i] = c->dsp->idct_permutation[z];
        c->lquant[p] = lquant[i];
        c->cquant[p] = cquant[i];
    }
    c->w = width;
    c->h = height;
}