This source file includes following definitions.
- y216_decode_init
- y216_decode_frame
#include "avcodec.h"
#include "internal.h"
static av_cold int y216_decode_init(AVCodecContext *avctx)
{
avctx->pix_fmt = AV_PIX_FMT_YUV422P16;
avctx->bits_per_raw_sample = 14;
return 0;
}
static int y216_decode_frame(AVCodecContext *avctx, void *data,
int *got_frame, AVPacket *avpkt)
{
AVFrame *pic = data;
const uint16_t *src = (uint16_t *)avpkt->data;
uint16_t *y, *u, *v, aligned_width = FFALIGN(avctx->width, 4);
int i, j, ret;
if (avpkt->size < 4 * avctx->height * aligned_width) {
av_log(avctx, AV_LOG_ERROR, "Insufficient input data.\n");
return AVERROR(EINVAL);
}
if ((ret = ff_get_buffer(avctx, pic, 0)) < 0)
return ret;
pic->key_frame = 1;
pic->pict_type = AV_PICTURE_TYPE_I;
y = (uint16_t *)pic->data[0];
u = (uint16_t *)pic->data[1];
v = (uint16_t *)pic->data[2];
for (i = 0; i < avctx->height; i++) {
for (j = 0; j < avctx->width >> 1; j++) {
u[ j ] = src[4 * j ] << 2 | src[4 * j ] >> 14;
y[2 * j ] = src[4 * j + 1] << 2 | src[4 * j + 1] >> 14;
v[ j ] = src[4 * j + 2] << 2 | src[4 * j + 2] >> 14;
y[2 * j + 1] = src[4 * j + 3] << 2 | src[4 * j + 3] >> 14;
}
y += pic->linesize[0] >> 1;
u += pic->linesize[1] >> 1;
v += pic->linesize[2] >> 1;
src += aligned_width << 1;
}
*got_frame = 1;
return avpkt->size;
}
AVCodec ff_targa_y216_decoder = {
.name = "targa_y216",
.long_name = NULL_IF_CONFIG_SMALL("Pinnacle TARGA CineWave YUV16"),
.type = AVMEDIA_TYPE_VIDEO,
.id = AV_CODEC_ID_TARGA_Y216,
.init = y216_decode_init,
.decode = y216_decode_frame,
.capabilities = CODEC_CAP_DR1,
};