* For more information on the OpenEXR format, visit:
* http://openexr.com/
*
- * exr_flt2uint() and exr_halflt2uint() is credited to Reimar Döffinger
+ * exr_flt2uint() and exr_halflt2uint() is credited to Reimar Döffinger.
+ * exr_half2float() is credited to Aaftab Munshi, Dan Ginsburg, Dave Shreiner.
*/
+#include <float.h>
#include <zlib.h>
#include "libavutil/imgutils.h"
+#include "libavutil/intfloat.h"
#include "libavutil/opt.h"
#include "avcodec.h"
+#include "bitstream.h"
#include "bytestream.h"
-#include "get_bits.h"
#include "internal.h"
#include "mathops.h"
#include "thread.h"
EXRThreadData *thread_data;
const char *layer;
+
+ float gamma;
+ uint16_t gamma_table[65536];
} EXRContext;
+/* -15 stored using a single precision bias of 127 */
+#define HALF_FLOAT_MIN_BIASED_EXP_AS_SINGLE_FP_EXP 0x38000000
+
+/* max exponent value in single precision that will be converted
+ * to Inf or Nan when stored as a half-float */
+#define HALF_FLOAT_MAX_BIASED_EXP_AS_SINGLE_FP_EXP 0x47800000
+
+/* 255 is the max exponent biased value */
+#define FLOAT_MAX_BIASED_EXP (0xFF << 23)
+
+#define HALF_FLOAT_MAX_BIASED_EXP (0x1F << 10)
+
+/**
+ * Convert a half float as a uint16_t into a full float.
+ *
+ * @param hf half float as uint16_t
+ *
+ * @return float value
+ */
+static union av_intfloat32 exr_half2float(uint16_t hf)
+{
+ unsigned int sign = (unsigned int) (hf >> 15);
+ unsigned int mantissa = (unsigned int) (hf & ((1 << 10) - 1));
+ unsigned int exp = (unsigned int) (hf & HALF_FLOAT_MAX_BIASED_EXP);
+ union av_intfloat32 f;
+
+ if (exp == HALF_FLOAT_MAX_BIASED_EXP) {
+ // we have a half-float NaN or Inf
+ // half-float NaNs will be converted to a single precision NaN
+ // half-float Infs will be converted to a single precision Inf
+ exp = FLOAT_MAX_BIASED_EXP;
+ if (mantissa)
+ mantissa = (1 << 23) - 1; // set all bits to indicate a NaN
+ } else if (exp == 0x0) {
+ // convert half-float zero/denorm to single precision value
+ if (mantissa) {
+ mantissa <<= 1;
+ exp = HALF_FLOAT_MIN_BIASED_EXP_AS_SINGLE_FP_EXP;
+ // check for leading 1 in denorm mantissa
+ while ((mantissa & (1 << 10))) {
+ // for every leading 0, decrement single precision exponent by 1
+ // and shift half-float mantissa value to the left
+ mantissa <<= 1;
+ exp -= (1 << 23);
+ }
+ // clamp the mantissa to 10 bits
+ mantissa &= ((1 << 10) - 1);
+ // shift left to generate single-precision mantissa of 23 bits
+ mantissa <<= 13;
+ }
+ } else {
+ // shift left to generate single-precision mantissa of 23 bits
+ mantissa <<= 13;
+ // generate single precision biased exponent value
+ exp = (exp << 13) + HALF_FLOAT_MIN_BIASED_EXP_AS_SINGLE_FP_EXP;
+ }
+
+ f.i = (sign << 31) | exp | mantissa;
+
+ return f;
+}
+
+
/**
* Convert from 32-bit float as uint32_t to uint16_t.
*
static int huf_unpack_enc_table(GetByteContext *gb,
int32_t im, int32_t iM, uint64_t *hcode)
{
- GetBitContext gbit;
-
- init_get_bits8(&gbit, gb->buffer, bytestream2_get_bytes_left(gb));
+ BitstreamContext bc;
+ int ret = bitstream_init8(&bc, gb->buffer, bytestream2_get_bytes_left(gb));
+ if (ret < 0)
+ return ret;
for (; im <= iM; im++) {
- uint64_t l = hcode[im] = get_bits(&gbit, 6);
+ uint64_t l = hcode[im] = bitstream_read(&bc, 6);
if (l == LONG_ZEROCODE_RUN) {
- int zerun = get_bits(&gbit, 8) + SHORTEST_LONG_RUN;
+ int zerun = bitstream_read(&bc, 8) + SHORTEST_LONG_RUN;
if (im + zerun > iM + 1)
return AVERROR_INVALIDDATA;
}
}
- bytestream2_skip(gb, (get_bits_count(&gbit) + 7) / 8);
+ bytestream2_skip(gb, (bitstream_tell(&bc) + 7) / 8);
huf_canonical_code_table(hcode);
return 0;
lc += 8; \
}
-#define get_code(po, rlc, c, lc, gb, out, oe) \
+#define get_code(po, rlc, c, lc, gb, out, oe, outb) \
{ \
if (po == rlc) { \
if (lc < 8) \
\
cs = c >> lc; \
\
- if (out + cs > oe) \
+ if (out + cs > oe || out == outb) \
return AVERROR_INVALIDDATA; \
\
s = out[-1]; \
if (pl.len) {
lc -= pl.len;
- get_code(pl.lit, rlc, c, lc, gb, out, oe);
+ get_code(pl.lit, rlc, c, lc, gb, out, oe, outb);
} else {
int j;
if ((hcode[pl.p[j]] >> 6) ==
((c >> (lc - l)) & ((1LL << l) - 1))) {
lc -= l;
- get_code(pl.p[j], rlc, c, lc, gb, out, oe);
+ get_code(pl.p[j], rlc, c, lc, gb, out, oe, outb);
break;
}
}
if (pl.len) {
lc -= pl.len;
- get_code(pl.lit, rlc, c, lc, gb, out, oe);
+ get_code(pl.lit, rlc, c, lc, gb, out, oe, outb);
} else {
return AVERROR_INVALIDDATA;
}
int axmax = (avctx->width - (s->xmax + 1)) * 2 * s->desc->nb_components;
int bxmin = s->xmin * 2 * s->desc->nb_components;
int i, x, buf_size = s->buf_size;
+ float one_gamma = 1.0f / s->gamma;
int ret;
line_offset = AV_RL64(s->gb.buffer + jobnr * 8);
if (s->pixel_type == EXR_FLOAT) {
// 32-bit
for (x = 0; x < xdelta; x++) {
- *ptr_x++ = exr_flt2uint(bytestream_get_le32(&r));
- *ptr_x++ = exr_flt2uint(bytestream_get_le32(&g));
- *ptr_x++ = exr_flt2uint(bytestream_get_le32(&b));
+ union av_intfloat32 t;
+ t.i = bytestream_get_le32(&r);
+ if (t.f > 0.0f) /* avoid negative values */
+ t.f = powf(t.f, one_gamma);
+ *ptr_x++ = exr_flt2uint(t.i);
+
+ t.i = bytestream_get_le32(&g);
+ if (t.f > 0.0f)
+ t.f = powf(t.f, one_gamma);
+ *ptr_x++ = exr_flt2uint(t.i);
+
+ t.i = bytestream_get_le32(&b);
+ if (t.f > 0.0f)
+ t.f = powf(t.f, one_gamma);
+ *ptr_x++ = exr_flt2uint(t.i);
if (channel_buffer[3])
*ptr_x++ = exr_flt2uint(bytestream_get_le32(&a));
}
} else {
// 16-bit
for (x = 0; x < xdelta; x++) {
- *ptr_x++ = exr_halflt2uint(bytestream_get_le16(&r));
- *ptr_x++ = exr_halflt2uint(bytestream_get_le16(&g));
- *ptr_x++ = exr_halflt2uint(bytestream_get_le16(&b));
+ *ptr_x++ = s->gamma_table[bytestream_get_le16(&r)];
+ *ptr_x++ = s->gamma_table[bytestream_get_le16(&g)];
+ *ptr_x++ = s->gamma_table[bytestream_get_le16(&b)];
if (channel_buffer[3])
*ptr_x++ = exr_halflt2uint(bytestream_get_le16(&a));
}
current_pixel_type = bytestream2_get_le32(&ch_gb);
if (current_pixel_type >= EXR_UNKNOWN) {
- avpriv_report_missing_feature(s->avctx,
- "Pixel type %d.\n",
+ avpriv_report_missing_feature(s->avctx, "Pixel type %d",
current_pixel_type);
return AVERROR_PATCHWELCOME;
}
static av_cold int decode_init(AVCodecContext *avctx)
{
EXRContext *s = avctx->priv_data;
+ uint32_t i;
+ union av_intfloat32 t;
+ float one_gamma = 1.0f / s->gamma;
s->avctx = avctx;
s->xmin = ~0;
s->w = 0;
s->h = 0;
- // allocate thread data, used for non EXR_RAW compreesion types
+ if (one_gamma > 0.9999f && one_gamma < 1.0001f) {
+ for (i = 0; i < 65536; ++i)
+ s->gamma_table[i] = exr_halflt2uint(i);
+ } else {
+ for (i = 0; i < 65536; ++i) {
+ t = exr_half2float(i);
+ /* If negative value we reuse half value */
+ if (t.f <= 0.0f) {
+ s->gamma_table[i] = exr_halflt2uint(i);
+ } else {
+ t.f = powf(t.f, one_gamma);
+ s->gamma_table[i] = exr_flt2uint(t.i);
+ }
+ }
+ }
+
+ // allocate thread data, used for non EXR_RAW compression types
s->thread_data = av_mallocz_array(avctx->thread_count, sizeof(EXRThreadData));
if (!s->thread_data)
return AVERROR_INVALIDDATA;
static int decode_init_thread_copy(AVCodecContext *avctx)
{ EXRContext *s = avctx->priv_data;
- // allocate thread data, used for non EXR_RAW compreesion types
+ // allocate thread data, used for non EXR_RAW compression types
s->thread_data = av_mallocz_array(avctx->thread_count, sizeof(EXRThreadData));
if (!s->thread_data)
return AVERROR_INVALIDDATA;
static const AVOption options[] = {
{ "layer", "Set the decoding layer", OFFSET(layer),
AV_OPT_TYPE_STRING, { .str = "" }, 0, 0, VD },
+ { "gamma", "Set the float gamma value when decoding", OFFSET(gamma),
+ AV_OPT_TYPE_FLOAT, { .dbl = 1.0f }, 0.001, FLT_MAX, VD },
{ NULL },
};
.init_thread_copy = ONLY_IF_THREADS_ENABLED(decode_init_thread_copy),
.close = decode_end,
.decode = decode_frame,
- .capabilities = CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS |
- CODEC_CAP_SLICE_THREADS,
+ .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS |
+ AV_CODEC_CAP_SLICE_THREADS,
.priv_class = &exr_class,
};