Signed-off-by: Paul B Mahol <one...@gmail.com> --- libavcodec/scpr.c | 1606 +++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 1568 insertions(+), 38 deletions(-)
diff --git a/libavcodec/scpr.c b/libavcodec/scpr.c index 965a3ed22f..bef6d10130 100644 --- a/libavcodec/scpr.c +++ b/libavcodec/scpr.c @@ -24,47 +24,1562 @@ #include <stdlib.h> #include <string.h> +#include "libavutil/qsort.h" + #include "avcodec.h" #include "bytestream.h" #include "internal.h" -#define TOP 0x01000000 -#define BOT 0x010000 +#define TOP 0x01000000 +#define BOT 0x010000 + +typedef struct RangeCoder { + unsigned code; + unsigned range; + unsigned code1; +} RangeCoder; + +typedef struct PixelModel { + unsigned freq[256]; + unsigned lookup[16]; + unsigned total_freq; +} PixelModel; + +typedef struct PixelModel3 { + uint8_t type; + uint8_t length; + uint8_t maxpos; + uint8_t fshift; + uint16_t size; + uint32_t cntsum; + uint8_t symbols[256]; + uint16_t freqs[256]; + uint16_t freqs1[256]; + uint16_t cnts[256]; + uint8_t dectab[32]; +} PixelModel3; + +typedef struct FillModel3 { + uint32_t cntsum; + uint16_t freqs[2][5]; + uint16_t cnts[5]; + uint8_t dectab[32]; +} FillModel3; + +typedef struct OpModel3 { + uint32_t cntsum; + uint16_t freqs[2][6]; + uint16_t cnts[6]; + uint8_t dectab[32]; +} OpModel3; + +typedef struct RunModel3 { + uint32_t cntsum; + uint16_t freqs[2][256]; + uint16_t cnts[256]; + uint8_t dectab[32]; +} RunModel3; + +typedef struct SxyModel3 { + uint32_t cntsum; + uint16_t freqs[2][16]; + uint16_t cnts[16]; + uint8_t dectab[32]; +} SxyModel3; + +typedef struct MVModel3 { + uint32_t cntsum; + uint16_t freqs[2][512]; + uint16_t cnts[512]; + uint8_t dectab[32]; +} MVModel3; + +typedef struct SCPRContext { + int version; + AVFrame *last_frame; + AVFrame *current_frame; + GetByteContext gb; + RangeCoder rc; + PixelModel pixel_model[3][4096]; + unsigned op_model[6][7]; + unsigned run_model[6][257]; + unsigned range_model[257]; + unsigned count_model[257]; + unsigned fill_model[6]; + unsigned sxy_model[4][17]; + unsigned mv_model[2][513]; + unsigned nbx, nby; + unsigned nbcount; + unsigned *blocks; + unsigned cbits; + int cxshift; + + PixelModel3 pixel_model3[3][4096]; + RunModel3 run_model3[6]; + RunModel3 range_model3; + RunModel3 count_model3; + FillModel3 fill_model3; + SxyModel3 sxy_model3[4]; + MVModel3 mv_model3[2]; + OpModel3 op_model3[6]; + + int (*get_freq)(RangeCoder *rc, unsigned total_freq, unsigned *freq); + int (*decode)(GetByteContext *gb, RangeCoder *rc, unsigned cumFreq, unsigned freq, unsigned total_freq); +} SCPRContext; + +static void renew_table3(unsigned nsym, uint32_t *cntsum, + uint16_t *freqs, uint16_t *freqs1, + uint16_t *cnts, uint8_t *dectab) +{ + unsigned a = 0, b = 4096 / nsym, c = b - (b >> 1); + + *cntsum = c * nsym; + + for (int d = 0; d < nsym; d++) { + freqs[d] = b; + freqs1[d] = a; + cnts[d] = c; + for (int q = a + 128 - 1 >> 7, f = (a + b - 1 >> 7) + 1; q < f; q++) { + dectab[q] = d; + } + + a += b; + } +} + +static void reinit_tables3(SCPRContext * s) +{ + int i, j; + + for (i = 0; i < 3; i++) { + for (j = 0; j < 4096; j++) { + PixelModel3 *m = &s->pixel_model3[i][j]; + memset(m, 0, sizeof(*m)); + } + } + + for (i = 0; i < 6; i++) { + renew_table3(256, &s->run_model3[i].cntsum, + s->run_model3[i].freqs[0], s->run_model3[i].freqs[1], + s->run_model3[i].cnts, s->run_model3[i].dectab); + } + + renew_table3(256, &s->range_model3.cntsum, + s->range_model3.freqs[0], s->range_model3.freqs[1], + s->range_model3.cnts, s->range_model3.dectab); + + renew_table3(5, &s->fill_model3.cntsum, + s->fill_model3.freqs[0], s->fill_model3.freqs[1], + s->fill_model3.cnts, s->fill_model3.dectab); + + renew_table3(256, &s->count_model3.cntsum, + s->count_model3.freqs[0], s->count_model3.freqs[1], + s->count_model3.cnts, s->count_model3.dectab); + + for (i = 0; i < 4; i++) { + renew_table3(16, &s->sxy_model3[i].cntsum, + s->sxy_model3[i].freqs[0], s->sxy_model3[i].freqs[1], + s->sxy_model3[i].cnts, s->sxy_model3[i].dectab); + } + + for (i = 0; i < 2; i++) { + renew_table3(512, &s->mv_model3[i].cntsum, + s->mv_model3[i].freqs[0], s->mv_model3[i].freqs[1], + s->mv_model3[i].cnts, s->mv_model3[i].dectab); + } + + for (i = 0; i < 6; i++) { + renew_table3(6, &s->op_model3[i].cntsum, + s->op_model3[i].freqs[0], s->op_model3[i].freqs[1], + s->op_model3[i].cnts, s->op_model3[i].dectab); + } +} + +static int decode3(GetByteContext *gb, RangeCoder *rc, unsigned a, unsigned b) +{ + unsigned x = a * (rc->code >> 12) + (rc->code & 0xFFF) - b; + + while (x < 0x800000 && bytestream2_get_bytes_left(gb) > 0) { + x = bytestream2_get_byte(gb) | (x << 8); + } + rc->code = x; + + return 0; +} + +static void rescale(PixelModel3 *m, int *totfr) +{ + unsigned a, b, c; + + for (a = 256 - m->size, b = 0, c = m->size; b < c; b++) { + m->freqs[b] -= m->freqs[b] >> 1; + a += m->freqs[b]; + } + + *totfr = a; +} + +static int add_symbol(PixelModel3 *m, int index, unsigned symbol, int *totfr, int max) +{ + if (m->size == max) + return 0; + + for (int c = m->size - 1; c >= index; c--) { + m->symbols[c + 1] = m->symbols[c]; + m->freqs[c + 1] = m->freqs[c]; + } + + m->symbols[index] = symbol; + m->freqs[index] = 50; + m->size++; + + if (m->maxpos >= index) + m->maxpos++; + + *totfr += 50; + if (*totfr + 50 > 4096) + rescale(m, totfr); + + return 1; +} + +static int decode_adaptive4(PixelModel3 *m, unsigned rccode, unsigned *value, + uint16_t *a, uint16_t *b, uint16_t c) +{ + unsigned q, g, maxpos, d, e = c, totfr = c; + + for (d = 0; e <= 2048; d++) + e <<= 1; + maxpos = m->maxpos; + rccode >>= d; + c = m->freqs[maxpos]; + m->freqs[maxpos] += 4096 - e >> d; + + for (q = 0, g = 0, e = 0; q < m->size; q++) { + unsigned f = m->symbols[q]; + unsigned p = e + f - g; + unsigned k = m->freqs[q]; + + if (rccode < p) { + *value = rccode - e + g; + *b = rccode << d; + *a = 1 << d; + m->freqs[maxpos] = c; + return add_symbol(m, q, *value, &totfr, 4); + } + + if (p + k > rccode) { + *value = f; + e += *value - g; + *b = e << d; + *a = k << d; + m->freqs[maxpos] = c; + m->freqs[q] += 50; + totfr += 50; + if ((q != maxpos) && (m->freqs[q] > m->freqs[maxpos])) + m->maxpos = q; + if (totfr + 50 > 4096) + rescale(m, &totfr); + return 1; + } + + e += f - g + k; + g = f + 1; + } + + m->freqs[maxpos] = c; + *value = g + rccode - e; + *b = rccode << d; + *a = 1 << d; + return add_symbol(m, q, *value, &totfr, 4); +} + +static int decode_adaptive5(PixelModel3 *m, int rccode, unsigned *value, + uint16_t *a, uint16_t *b, uint32_t *c) +{ + unsigned q, g, maxpos, d, e = *c, totfr = *c; + int ret; + + for (d = 0; e <= 2048; d++) + e <<= 1; + maxpos = m->maxpos; + rccode >>= d; + *c = m->freqs[maxpos]; + m->freqs[maxpos] += 4096 - e >> d; + + for (q = 0, g = 0, e = 0; q < m->size; q++) { + unsigned f = m->symbols[q]; + unsigned p = e + f - g; + unsigned k = m->freqs[q]; + + if (rccode < p) { + *value = rccode - e + g; + *b = rccode << d; + *a = 1 << d; + m->freqs[maxpos] = *c; + ret = add_symbol(m, q, *value, &totfr, 16); + *c = totfr; + return ret; + } + + if (p + k > rccode) { + *value = f; + e += *value - g; + *b = e << d; + *a = k << d; + m->freqs[maxpos] = *c; + m->freqs[q] += 50; + totfr += 50; + if ((q != maxpos) && (m->freqs[q] > m->freqs[maxpos])) + m->maxpos = q; + if (totfr + 50 > 4096) + rescale(m, &totfr); + *c = totfr; + return 1; + } + + e += f - g + k; + g = f + 1; + } + + m->freqs[maxpos] = *c; + *value = g + rccode - e; + *b = rccode << d; + *a = 1 << d; + ret = add_symbol(m, q, *value, &totfr, 16); + *c = totfr; + return ret; +} + +static int update_model6_to_7(PixelModel3 *n, PixelModel3 *m) +{ + int c, d, e, f, k, p, length, i, j, index; + uint16_t *freqs, *freqs1, *cnts; + + n->type = 7; + + length = m->length; + freqs = n->freqs; + freqs1 = n->freqs1; + cnts = n->cnts; + n->cntsum = m->cnts[length]; + for (i = 0; i < length; i++) { + if (!m->cnts[i]) + continue; + index = m->symbols[i]; + freqs[index] = m->freqs[2 * i]; + freqs1[index] = m->freqs[2 * i + 1]; + cnts[index] = m->cnts[i]; + } + c = 1 << m->fshift; + d = c - (c >> 1); + for (j = 0, e = 0; j < 256; j++) { + f = freqs[j]; + if (!f) { + f = c; + freqs[j] = c; + freqs1[j] = e; + cnts[j] = d; + } + p = (e + 127) >> 7; + k = ((f + e - 1) >> 7) + 1; + for (i = 0; i < k - p; i++) + n->dectab[p + i] = j; + e += f; + } + + memcpy(m, n, sizeof(*n)); + + return 0; +} + +static void calc_sum(PixelModel3 *m) +{ + int a, b, c; + + b = m->length; + a = 256 - m->size << (0 < m->fshift ? m->fshift - 1 : 0); + for (c = 0; c < b; c++) + a += m->cnts[c]; + m->cnts[b] = a; +} + +static void rescale_dec(PixelModel3 *m) +{ + uint16_t cnts[256] = {0}; + uint16_t freqs[512] = {0}; + int a, b, c, e, g; + + for (a = 1 << (0 < m->fshift ? m->fshift - 1 : 0), b = 0; b < 256; b++) { + cnts[b] = a; + } + + for (a = 0, b = m->size; a < b; a++) { + cnts[m->symbols[a]] = m->cnts[a]; + } + + for (b = a = 0; b < 256; b++) { + freqs[2 * b] = cnts[b]; + freqs[2 * b + 1] = a; + a += cnts[b]; + } + + if (m->fshift > 0) + m->fshift--; + + a = 256 - m->size << (0 < m->fshift ? m->fshift - 1 : 0); + for (b = 0, c = m->size; b < c; b++) { + m->cnts[b] -= m->cnts[b] >> 1; + a = a + m->cnts[b]; + e = m->symbols[b]; + g = freqs[2 * e + 1]; + m->freqs[2 * b] = freqs[2 * e]; + m->freqs[2 * b + 1] = g; + } + m->cnts[m->length] = a; +} + +static int update_model5_to_6(PixelModel3 *n, PixelModel3 *m, uint8_t value) +{ + int c, d, e, f, g, k, q, p; + + n->type = 6; + n->length = 32; + + for (c = m->size, d = 256 - c, e = 0; e < c; e++) + d = d + m->freqs[e]; + + for (e = 0; 2048 >= d; e++) + d <<= 1; + + for (q = d = 0, g = q = 0; g < c; g++) { + p = m->symbols[g]; + d = d + (p - q); + q = m->freqs[g]; + k = q << e; + n->freqs[2 * g] = k; + n->freqs[2 * g + 1] = d << e; + n->cnts[g] = k - (k >> 1); + n->symbols[g] = p; + d += q; + q = p + 1; + } + + n->fshift = e; + e = 1 << n->fshift; + d = 0; + if (value > 0) { + d = -1; + for (p = f = g = 0; p < c; p++) { + k = n->symbols[p]; + if (k > d && k < value) { + d = k; + g = n->freqs[2 * p]; + f = n->freqs[2 * p + 1]; + } + } + d = 0 < g ? f + g + (value - d - 1 << n->fshift) : value << n->fshift; + } + n->freqs[2 * c] = e; + n->freqs[2 * c + 1] = d; + n->cnts[c] = e - (e >> 1); + n->symbols[c] = value; + n->size = c + 1; + e = 25 << n->fshift; + n->cnts[c] += e; + n->cnts[32] += e; + if (n->cnts[32] + e > 4096) { + rescale_dec(n); + } + calc_sum(n); + c = 0; + for (e = n->size - 1; c < e;) { + for (d = c++, g = d + 1, f = n->size; g < f;) { + if (p = g++, q = n->freqs[2 * p], k = n->freqs[2 * d], q > k) { + int l = n->freqs[2 * d + 1]; + int h = n->freqs[2 * p + 1]; + n->freqs[2 * d] = q; + n->freqs[2 * d + 1] = h; + n->freqs[2 * p] = k; + n->freqs[2 * p + 1] = l; + q = n->cnts[d]; + n->cnts[d] = n->cnts[p]; + n->cnts[p] = q; + q = n->symbols[d]; + n->symbols[d] = n->symbols[p]; + n->symbols[p] = q; + } + } + } + + memcpy(m, n, sizeof(*n)); + + return 0; +} + +static void grow_dec(PixelModel3 *m) +{ + int a; + + a = 2 * m->length; + m->cnts[2 * m->length] = m->cnts[m->length]; + m->length = a; +} + +static int add_dec(PixelModel3 *m, int a, int b, int c) +{ + int d; + + if (m->size >= 40 || m->size >= m->length) + return -1; + + d = m->size; + m->symbols[d] = a; + m->freqs[2 * d] = b; + m->freqs[2 * d + 1] = c; + m->cnts[d] = b - (b >> 1); + m->size++; + + return d; +} + +static void incr_cntdec(PixelModel3 *m, int a) +{ + int b, c, d, e, g; + + b = 25 << m->fshift; + c = m->length; + m->cnts[a] += b; + m->cnts[c] += b; + if (a > 0 && m->cnts[a] > m->cnts[a - 1]) { + d = m->cnts[a]; + m->cnts[a] = m->cnts[a - 1]; + m->cnts[a - 1] = d; + d = m->freqs[2 * a]; + e = m->freqs[2 * a + 1]; + g = m->freqs[2 * (a - 1) + 1]; + m->freqs[2 * a] = m->freqs[2 * (a - 1)]; + m->freqs[2 * a + 1] = g; + g = a - 1; + m->freqs[2 * g] = d; + m->freqs[2 * g + 1] = e; + d = m->symbols[a]; + m->symbols[a] = m->symbols[a - 1]; + m->symbols[a - 1] = d; + } + if (m->cnts[c] + b > 4096) + rescale_dec(m); +} + +static int decode_adaptive6(PixelModel3 *m, unsigned code, unsigned *value, + uint16_t *a, uint16_t *b) +{ + int c, d, e, f, g, q; + + for (c = 0, d = 0, e = 0, f = 0, g = 0, q = m->size; g < q; g++) { + unsigned p = m->freqs[2 * g + 1]; + + if (p <= code) { + unsigned k = m->freqs[2 * g]; + + if (p + k > code) { + *value = m->symbols[g]; + *a = k; + *b = p; + incr_cntdec(m, g); + return 1; + } + + if (p >= d) { + c = k; + d = p; + e = m->symbols[g]; + } + } + } + + g = 1 << m->fshift; + q = f = 0; + + if (c > 0) { + f = code - (d + c) >> m->fshift; + q = f + e + 1; + f = d + c + (f << m->fshift); + } else { + q = code >> m->fshift; + f = q << m->fshift; + } + + *a = g; + *b = f; + *value = q; + + c = add_dec(m, q, g, f); + if (c < 0) { + if (m->length == 64) + return 0; + grow_dec(m); + c = add_dec(m, q, g, f); + } + + incr_cntdec(m, c); + return 1; +} + +static int cmpbytes(const void *p1, const void *p2) +{ + int left = *(const uint8_t *)p1; + int right = *(const uint8_t *)p2; + return FFDIFFSIGN(left, right); +} + +static int update_model1_to_2(PixelModel3 *m, unsigned val) +{ + PixelModel3 n = {0}; + int i, b; + + n.type = 2; + n.size = m->size + 1; + b = m->size; + for (i = 0; i < b; i++) + n.symbols[i] = m->symbols[i]; + n.symbols[b] = val; + + memcpy(m, &n, sizeof(n)); + + return 0; +} + +static int update_model1_to_4(PixelModel3 *m, unsigned val) +{ + PixelModel3 n = {0}; + int size, i; + + size = m->size; + n.type = 4; + n.size = size; + for (i = 0; i < n.size; i++) { + n.symbols[i] = m->symbols[i]; + } + AV_QSORT(n.symbols, size, uint8_t, cmpbytes); + for (i = 0; i < n.size; i++) { + if (val == n.symbols[i]) { + n.freqs[i] = 100; + n.maxpos = i; + } else { + n.freqs[i] = 50; + } + } + + memcpy(m, &n, sizeof(n)); + + return 0; +} + +static int update_model1_to_5(PixelModel3 *m, unsigned val) +{ + PixelModel3 n = {0}; + int i, size, freqs; + unsigned a; + + size = m->size; + n.size = size; + for (i = 0; i < size; i++) { + n.symbols[i] = m->symbols[i]; + } + AV_QSORT(n.symbols, size, uint8_t, cmpbytes); + size = n.size; + for (i = 0; i < size; i++) { + if (val == n.symbols[i]) { + n.freqs[i] = 100; + n.maxpos = i; + } else { + n.freqs[i] = 50; + } + } + a = 256 - size; + for (i = 0; i < size; i++, a += freqs) + freqs = n.freqs[i]; + n.type = 5; + n.cntsum = a; + + memcpy(m, &n, sizeof(n)); + + return 0; +} + +static int decode_static1(PixelModel3 *m, unsigned val) +{ + unsigned size; + + size = m->size; + for (int i = 0; i < size; i++) { + if (val == m->symbols[i]) { + if (size <= 4) + return update_model1_to_4(m, val); + else + return update_model1_to_5(m, val); + } + } + + if (size >= 14) + return update_model1_to_2(m, val); + + m->symbols[size] = val; + m->size++; + return 0; +} + +static int update_model2_to_6(PixelModel3 *m, uint8_t value, int a4) +{ + PixelModel3 n = {0}; + int c, d, e, f, g, q; + + n.type = 6; + n.length = a4; + + memset(n.symbols, 1u, a4); + + c = m->size; + d = 256 - c + (64 * c + 64); + for (e = 0; d <= 2048; e++) { + d <<= 1; + } + + g = q = 0; + AV_QSORT(m->symbols, c, uint8_t, cmpbytes); + for (f = d = 0; f < c; f++) { + int p = f; + int k = m->symbols[p]; + int l; + g = g + (k - q); + + if (k == value) { + d = p; + q = 128; + } else { + q = 64; + } + l = q << e; + n.freqs[2 * p] = l; + n.freqs[2 * p + 1] = g << e; + n.symbols[p] = k; + n.cnts[p] = l - (l >> 1); + g += q; + q = k + 1; + } + n.size = c; + n.fshift = e; + calc_sum(&n); + + if (d > 0) { + c = n.freqs[0]; + e = n.freqs[1]; + g = n.freqs[2 * d + 1]; + n.freqs[0] = n.freqs[2 * d]; + n.freqs[1] = g; + n.freqs[2 * d] = c; + n.freqs[2 * d + 1] = e; + c = n.symbols[0]; + e = n.cnts[0]; + n.cnts[0] = n.cnts[d]; + n.cnts[d] = e; + n.symbols[0] = value; + n.symbols[d] = c; + } + + memcpy(m, &n, sizeof(n)); + + return 0; +} + +static int update_model2_to_3(PixelModel3 *m, unsigned val) +{ + PixelModel3 n = {0}; + unsigned size; + + n.type = 3; + n.size = m->size + 1; + + size = m->size; + for (int i = 0; i < size; i++) + n.symbols[i] = m->symbols[i]; + n.symbols[size] = val; + + memcpy(m, &n, sizeof(n)); + + return 0; +} + +static int decode_static2(PixelModel3 *m, unsigned val) +{ + unsigned size; + + size = m->size; + for (int i = 0; i < size; i++) { + if (val == m->symbols[i]) { + int a; + + if (m->size <= 32) + a = 32; + else + a = 64; + return update_model2_to_6(m, val, a); + } + } + + if (size >= 64) + return update_model2_to_3(m, val); + + m->symbols[size] = val; + m->size++; + + return 0; +} + +static int update_model3_to_7(PixelModel3 *m, uint8_t value) +{ + PixelModel3 n = {0}; + int c, d, e, f, g, q; + + n.type = 7; + + for (c = 0; c < 256; c++) { + d = c; + n.freqs[d] = 1; + n.cnts[d] = 1; + } + + for (c = m->size, d = (4096 - (256 - c)) / (c + 1) | 0, e = d - (d >> 1), g = 0; g < c;) { + q = g++; + q = m->symbols[q]; + n.freqs[q] = d; + n.cnts[q] = e; + } + n.freqs[value] += d; + n.cnts[value] += 16; + for (d = c = n.cntsum = 0; 256 > d; d++) { + e = d; + n.cntsum += n.cnts[e]; + n.freqs1[e] = c; + for (g = n.freqs[e], q = c + 128 - 1 >> 7, f = (c + g - 1 >> 7) + 1; q < f; q++) { + n.dectab[q] = e; + } + c += g; + } + + memcpy(m, &n, sizeof(n)); + + return 0; +} + +static int decode_static3(PixelModel3 *m, unsigned val) +{ + unsigned size = m->size; + + for (int i = 0; i < size; i++) { + if (val == m->symbols[i]) + return update_model3_to_7(m, val); + } + + if (size >= 256) + return 0; + + m->symbols[size] = val; + m->size++; + return 0; +} + +static void sync_code3(GetByteContext *gb, RangeCoder *rc) +{ + rc->code1++; + if (rc->code1 == 0x20000) { + rc->code = bytestream2_get_le32(gb); + rc->code1 = 0; + } +} + +static int decode_value3(SCPRContext *s, unsigned max, uint32_t *cntsum, + uint16_t *freqs1, uint16_t *freqs2, + uint16_t *cnts, uint8_t *dectable, + unsigned *value) +{ + GetByteContext *gb = &s->gb; + RangeCoder *rc = &s->rc; + unsigned r, y, a, b, c, e, g, q; + int i; + + r = dectable[(rc->code & 0xFFFu) >> 7]; + if (r < max) { + while (freqs2[r + 1] <= (rc->code & 0xFFF)) { + if (++r >= max) + break; + } + } + + cnts[r] += 16; + a = freqs1[r]; + b = freqs2[r]; + *cntsum += 16; + if (*cntsum + 16 > 4096) { + *cntsum = 0; + for (c = 0, i = 0; i < max + 1; i++) { + e = cnts[i]; + freqs2[i] = c; + freqs1[i] = e; + g = (c + 127) >> 7; + c += e; + q = ((c - 1) >> 7) + 1; + if (q > g) { + for (int j = 0; j < q - g; j++) + dectable[j + g] = i; + } + y = e - (e >> 1); + cnts[i] = y; + *cntsum += y; + } + } + + decode3(gb, rc, a, b); + sync_code3(gb, rc); + + *value = r; + + return 0; +} + +static void calc_sum5(PixelModel3 *m) +{ + int a, b, c; + + for (a = 256 - m->size, b = 0, c = m->size; b < c; b++) { + a += m->freqs[b]; + } + m->cntsum = a; +} + +static int update_model4_to_5(PixelModel3 *m, unsigned value) +{ + PixelModel3 n = {0}; + int c, d, e, g, totfr; + + n.type = 5; + + for (c = 0, d = m->size, e = 0; c < d && m->symbols[c] < value; c++) { + n.symbols[c] = m->symbols[c]; + e += n.freqs[c] = m->freqs[c]; + } + + g = c; + n.symbols[g] = value; + e += n.freqs[g] = 50; + for (++g; c < d;) { + n.symbols[g] = m->symbols[c]; + e += n.freqs[g] = m->freqs[c]; + ++c; + ++g; + } + n.size = d + 1; + if (e > 4096) + rescale(&n, &totfr); + + calc_sum5(&n); + + memcpy(m, &n, sizeof(n)); + + return 0; +} + +static signed int decode_unit3(SCPRContext *s, PixelModel3 *m, unsigned code, unsigned *value) +{ + PixelModel3 n = {0}; + GetByteContext *gb = &s->gb; + RangeCoder *rc = &s->rc; + uint16_t param, a = 0, b = 0; + int type; + + type = m->type; + switch (type) { + case 0: + *value = bytestream2_get_byte(&s->gb); + m->type = 1; + m->size = 1; + m->symbols[0] = *value; + sync_code3(gb, rc); + break; + case 1: + *value = bytestream2_get_byte(&s->gb); + decode_static1(m, *value); + sync_code3(gb, rc); + break; + case 2: + *value = bytestream2_get_byte(&s->gb); + decode_static2(m, *value); + sync_code3(gb, rc); + break; + case 3: + *value = bytestream2_get_byte(&s->gb); + decode_static3(m, *value); + sync_code3(gb, rc); + break; + case 4: + param = m->freqs[0] + m->freqs[1] + m->freqs[2] + m->freqs[3] + 256 - m->size; + if (!decode_adaptive4(m, code, value, &a, &b, param)) { + update_model4_to_5(m, *value); + } + decode3(gb, rc, a, b); + sync_code3(gb, rc); + break; + case 5: + if (!decode_adaptive5(m, code, value, &a, &b, &m->cntsum)) { + update_model5_to_6(&n, m, *value); + } + decode3(gb, rc, a, b); + sync_code3(gb, rc); + break; + case 6: + if (!decode_adaptive6(m, code, value, &a, &b)) { + update_model6_to_7(&n, m); + } + decode3(gb, rc, a, b); + sync_code3(gb, rc); + break; + case 7: + return decode_value3(s, 255, &m->cntsum, + m->freqs, m->freqs1, + m->cnts, m->dectab, value); + } + return 0; +} + +static int decode_units3(SCPRContext * s, unsigned *red, + unsigned *green, unsigned *blue, + int *cx, int *cx1) +{ + RangeCoder *rc = &s->rc; + + decode_unit3(s, &s->pixel_model3[0][*cx + *cx1], rc->code & 0xFFF, red); + + *cx1 = (*cx << 6) & 0xFC0; + *cx = *red >> 2; + + decode_unit3(s, &s->pixel_model3[1][*cx + *cx1], rc->code & 0xFFF, green); + + *cx1 = (*cx << 6) & 0xFC0; + *cx = *green >> 2; + + decode_unit3(s, &s->pixel_model3[2][*cx + *cx1], rc->code & 0xFFF, blue); + + *cx1 = (*cx << 6) & 0xFC0; + *cx = *blue >> 2; + + return 0; +} + +static void init_rangecoder3(RangeCoder *rc, GetByteContext *gb) +{ + rc->code = bytestream2_get_le32(gb); + rc->code1 = 0; +} + +static int decompress_i3(AVCodecContext *avctx, uint32_t *dst, int linesize) +{ + SCPRContext *s = avctx->priv_data; + GetByteContext *gb = &s->gb; + RangeCoder *rc = &s->rc; + int cx = 0, cx1 = 0, k = 0, clr = 0; + int run, off, y = 0, x = 0, z, ret; + unsigned backstep = linesize - avctx->width; + unsigned lx, ly, ptype, r, g, b; + + bytestream2_skip(gb, 1); + init_rangecoder3(rc, gb); + reinit_tables3(s); + + while (k < avctx->width + 1) { + decode_units3(s, &r, &g, &b, &cx, &cx1); + ret = decode_value3(s, 255, &s->run_model3[0].cntsum, + s->run_model3[0].freqs[0], + s->run_model3[0].freqs[1], + s->run_model3[0].cnts, + s->run_model3[0].dectab, &run); + if (ret < 0) + return ret; + if (run <= 0) + return AVERROR_INVALIDDATA; + + clr = (b << 16) + (g << 8) + r; + k += run; + while (run-- > 0) { + if (y >= avctx->height) + return AVERROR_INVALIDDATA; + + dst[y * linesize + x] = clr; + lx = x; + ly = y; + x++; + if (x >= avctx->width) { + x = 0; + y++; + } + } + } + off = -linesize - 1; + ptype = 0; + + while (x < avctx->width && y < avctx->height) { + ret = decode_value3(s, 5, &s->op_model3[ptype].cntsum, + s->op_model3[ptype].freqs[0], + s->op_model3[ptype].freqs[1], + s->op_model3[ptype].cnts, + s->op_model3[ptype].dectab, &ptype); + if (ret < 0) + return ret; + if (ptype == 0) { + decode_units3(s, &r, &g, &b, &cx, &cx1); + clr = (b << 16) + (g << 8) + r; + } + if (ptype > 5) + return AVERROR_INVALIDDATA; + ret = decode_value3(s, 255, &s->run_model3[ptype].cntsum, + s->run_model3[ptype].freqs[0], + s->run_model3[ptype].freqs[1], + s->run_model3[ptype].cnts, + s->run_model3[ptype].dectab, &run); + if (ret < 0) + return ret; + if (run <= 0) + return AVERROR_INVALIDDATA; + + switch (ptype) { + case 0: + while (run-- > 0) { + if (y >= avctx->height) + return AVERROR_INVALIDDATA; + + dst[y * linesize + x] = clr; + lx = x; + ly = y; + x++; + if (x >= avctx->width) { + x = 0; + y++; + } + } + break; + case 1: + while (run-- > 0) { + if (y >= avctx->height) + return AVERROR_INVALIDDATA; + + dst[y * linesize + x] = dst[ly * linesize + lx]; + lx = x; + ly = y; + x++; + if (x >= avctx->width) { + x = 0; + y++; + } + } + clr = dst[ly * linesize + lx]; + break; + case 2: + while (run-- > 0) { + if (y < 1 || y >= avctx->height) + return AVERROR_INVALIDDATA; + + clr = dst[y * linesize + x + off + 1]; + dst[y * linesize + x] = clr; + lx = x; + ly = y; + x++; + if (x >= avctx->width) { + x = 0; + y++; + } + } + break; + case 4: + while (run-- > 0) { + uint8_t *odst = (uint8_t *)dst; + + if (y < 1 || y >= avctx->height || + (y == 1 && x == 0)) + return AVERROR_INVALIDDATA; + + if (x == 0) { + z = backstep; + } else { + z = 0; + } + + r = odst[(ly * linesize + lx) * 4] + + odst[((y * linesize + x) + off) * 4 + 4] - + odst[((y * linesize + x) + off - z) * 4]; + g = odst[(ly * linesize + lx) * 4 + 1] + + odst[((y * linesize + x) + off) * 4 + 5] - + odst[((y * linesize + x) + off - z) * 4 + 1]; + b = odst[(ly * linesize + lx) * 4 + 2] + + odst[((y * linesize + x) + off) * 4 + 6] - + odst[((y * linesize + x) + off - z) * 4 + 2]; + clr = ((b & 0xFF) << 16) + ((g & 0xFF) << 8) + (r & 0xFF); + dst[y * linesize + x] = clr; + lx = x; + ly = y; + x++; + if (x >= avctx->width) { + x = 0; + y++; + } + } + break; + case 5: + while (run-- > 0) { + if (y < 1 || y >= avctx->height || + (y == 1 && x == 0)) + return AVERROR_INVALIDDATA; + + if (x == 0) { + z = backstep; + } else { + z = 0; + } + + clr = dst[y * linesize + x + off - z]; + dst[y * linesize + x] = clr; + lx = x; + ly = y; + x++; + if (x >= avctx->width) { + x = 0; + y++; + } + } + break; + } + + if (avctx->bits_per_coded_sample == 16) { + cx1 = (clr & 0x3F00) >> 2; + cx = (clr & 0x3FFFFF) >> 16; + } else { + cx1 = (clr & 0xFC00) >> 4; + cx = (clr & 0xFFFFFF) >> 18; + } + } -typedef struct RangeCoder { - unsigned code; - unsigned range; - unsigned code1; -} RangeCoder; + return 0; +} -typedef struct PixelModel { - unsigned freq[256]; - unsigned lookup[16]; - unsigned total_freq; -} PixelModel; +static int decompress_p3(AVCodecContext *avctx, + uint32_t *dst, int linesize, + uint32_t *prev, int plinesize) +{ + SCPRContext *s = avctx->priv_data; + GetByteContext *gb = &s->gb; + int ret, temp, min, max, x, y, cx = 0, cx1 = 0; + int backstep = linesize - avctx->width; + int mvx = 0, mvy = 0; -typedef struct SCPRContext { - AVFrame *last_frame; - AVFrame *current_frame; - GetByteContext gb; - RangeCoder rc; - PixelModel pixel_model[3][4096]; - unsigned op_model[6][7]; - unsigned run_model[6][257]; - unsigned range_model[257]; - unsigned count_model[257]; - unsigned fill_model[6]; - unsigned sxy_model[4][17]; - unsigned mv_model[2][513]; - unsigned nbx, nby; - unsigned nbcount; - unsigned *blocks; - unsigned cbits; - int cxshift; + if (bytestream2_get_byte(gb) == 0) + return 0; + init_rangecoder3(&s->rc, gb); + + ret = decode_value3(s, 255, &s->range_model3.cntsum, + s->range_model3.freqs[0], + s->range_model3.freqs[1], + s->range_model3.cnts, + s->range_model3.dectab, &min); + ret |= decode_value3(s, 255, &s->range_model3.cntsum, + s->range_model3.freqs[0], + s->range_model3.freqs[1], + s->range_model3.cnts, + s->range_model3.dectab, &temp); + min += temp << 8; + ret |= decode_value3(s, 255, &s->range_model3.cntsum, + s->range_model3.freqs[0], + s->range_model3.freqs[1], + s->range_model3.cnts, + s->range_model3.dectab, &max); + ret |= decode_value3(s, 255, &s->range_model3.cntsum, + s->range_model3.freqs[0], + s->range_model3.freqs[1], + s->range_model3.cnts, + s->range_model3.dectab, &temp); + if (ret < 0) + return ret; - int (*get_freq)(RangeCoder *rc, unsigned total_freq, unsigned *freq); - int (*decode)(GetByteContext *gb, RangeCoder *rc, unsigned cumFreq, unsigned freq, unsigned total_freq); -} SCPRContext; + max += temp << 8; + memset(s->blocks, 0, sizeof(*s->blocks) * s->nbcount); + + while (min <= max) { + int fill, count; + + ret = decode_value3(s, 4, &s->fill_model3.cntsum, + s->fill_model3.freqs[0], + s->fill_model3.freqs[1], + s->fill_model3.cnts, + s->fill_model3.dectab, &fill); + ret |= decode_value3(s, 255, &s->count_model3.cntsum, + s->count_model3.freqs[0], + s->count_model3.freqs[1], + s->count_model3.cnts, + s->count_model3.dectab, &count); + if (ret < 0) + return ret; + + while (min < s->nbcount && count-- > 0) { + s->blocks[min++] = fill; + } + } + + for (y = 0; y < s->nby; y++) { + for (x = 0; x < s->nbx; x++) { + int sy1 = 0, sy2 = 16, sx1 = 0, sx2 = 16; + + if (s->blocks[y * s->nbx + x] == 0) + continue; + + if (((s->blocks[y * s->nbx + x] + 1) & 1) > 0) { + ret = decode_value3(s, 15, &s->sxy_model3[0].cntsum, + s->sxy_model3[0].freqs[0], + s->sxy_model3[0].freqs[1], + s->sxy_model3[0].cnts, + s->sxy_model3[0].dectab, &sx1); + ret |= decode_value3(s, 15, &s->sxy_model3[1].cntsum, + s->sxy_model3[1].freqs[0], + s->sxy_model3[1].freqs[1], + s->sxy_model3[1].cnts, + s->sxy_model3[1].dectab, &sy1); + ret |= decode_value3(s, 15, &s->sxy_model3[2].cntsum, + s->sxy_model3[2].freqs[0], + s->sxy_model3[2].freqs[1], + s->sxy_model3[2].cnts, + s->sxy_model3[2].dectab, &sx2); + ret |= decode_value3(s, 15, &s->sxy_model3[3].cntsum, + s->sxy_model3[3].freqs[0], + s->sxy_model3[3].freqs[1], + s->sxy_model3[3].cnts, + s->sxy_model3[3].dectab, &sy2); + if (ret < 0) + return ret; + + sx2++; + sy2++; + } + if (((s->blocks[y * s->nbx + x] + 3) & 2) > 0) { + int i, a, b, c, j, by = y * 16, bx = x * 16; + unsigned X; + + a = s->rc.code & 0xFFF; + c = 1; + + if (a < 0x800) + c = 0; + b = 2048; + if (!c) + b = 0; + + X = a + ((s->rc.code >> 1) & 0xFFFFF800) - b; + while (X < 0x800000 && bytestream2_get_bytes_left(gb) > 0) { + X = bytestream2_get_byte(gb) | (X << 8); + } + s->rc.code = X; + + sync_code3(gb, &s->rc); + + if (!c) { + ret = decode_value3(s, 511, &s->mv_model3[0].cntsum, + s->mv_model3[0].freqs[0], + s->mv_model3[0].freqs[1], + s->mv_model3[0].cnts, + s->mv_model3[0].dectab, &mvx); + ret |= decode_value3(s, 511, &s->mv_model3[1].cntsum, + s->mv_model3[1].freqs[0], + s->mv_model3[1].freqs[1], + s->mv_model3[1].cnts, + s->mv_model3[1].dectab, &mvy); + if (ret < 0) + return ret; + + mvx -= 256; + mvy -= 256; + } + + if (by + mvy + sy1 < 0 || bx + mvx + sx1 < 0 || + by + mvy + sy1 >= avctx->height || bx + mvx + sx1 >= avctx->width) + return AVERROR_INVALIDDATA; + + for (i = 0; i < sy2 - sy1 && (by + sy1 + i) < avctx->height && (by + mvy + sy1 + i) < avctx->height; i++) { + for (j = 0; j < sx2 - sx1 && (bx + sx1 + j) < avctx->width && (bx + mvx + sx1 + j) < avctx->width; j++) { + dst[(by + i + sy1) * linesize + bx + sx1 + j] = prev[(by + mvy + sy1 + i) * plinesize + bx + sx1 + mvx + j]; + } + } + } else { + int run, z, bx = x * 16 + sx1, by = y * 16 + sy1; + unsigned clr, ptype = 0, r, g, b; + + for (; by < y * 16 + sy2 && by < avctx->height;) { + ret = decode_value3(s, 5, &s->op_model3[ptype].cntsum, + s->op_model3[ptype].freqs[0], + s->op_model3[ptype].freqs[1], + s->op_model3[ptype].cnts, + s->op_model3[ptype].dectab, &ptype); + if (ret < 0) + return ret; + if (ptype == 0) { + ret = decode_units3(s, &r, &g, &b, &cx, &cx1); + if (ret < 0) + return ret; + + clr = (b << 16) + (g << 8) + r; + } + if (ptype > 5) + return AVERROR_INVALIDDATA; + ret = decode_value3(s, 255, &s->run_model3[ptype].cntsum, + s->run_model3[ptype].freqs[0], + s->run_model3[ptype].freqs[1], + s->run_model3[ptype].cnts, + s->run_model3[ptype].dectab, &run); + if (ret < 0) + return ret; + if (run <= 0) + return AVERROR_INVALIDDATA; + + switch (ptype) { + case 0: + while (run-- > 0) { + if (by >= avctx->height) + return AVERROR_INVALIDDATA; + + dst[by * linesize + bx] = clr; + bx++; + if (bx >= x * 16 + sx2 || bx >= avctx->width) { + bx = x * 16 + sx1; + by++; + } + } + break; + case 1: + while (run-- > 0) { + if (bx == 0) { + if (by < 1) + return AVERROR_INVALIDDATA; + z = backstep; + } else { + z = 0; + } + + if (by >= avctx->height) + return AVERROR_INVALIDDATA; + + clr = dst[by * linesize + bx - 1 - z]; + dst[by * linesize + bx] = clr; + bx++; + if (bx >= x * 16 + sx2 || bx >= avctx->width) { + bx = x * 16 + sx1; + by++; + } + } + break; + case 2: + while (run-- > 0) { + if (by < 1 || by >= avctx->height) + return AVERROR_INVALIDDATA; + + clr = dst[(by - 1) * linesize + bx]; + dst[by * linesize + bx] = clr; + bx++; + if (bx >= x * 16 + sx2 || bx >= avctx->width) { + bx = x * 16 + sx1; + by++; + } + } + break; + case 3: + while (run-- > 0) { + if (by >= avctx->height) + return AVERROR_INVALIDDATA; + + clr = prev[by * plinesize + bx]; + dst[by * linesize + bx] = clr; + bx++; + if (bx >= x * 16 + sx2 || bx >= avctx->width) { + bx = x * 16 + sx1; + by++; + } + } + break; + case 4: + while (run-- > 0) { + uint8_t *odst = (uint8_t *)dst; + + if (by < 1 || by >= avctx->height) + return AVERROR_INVALIDDATA; + + if (bx == 0) { + if (by < 2) + return AVERROR_INVALIDDATA; + z = backstep; + } else { + z = 0; + } + + r = odst[((by - 1) * linesize + bx) * 4] + + odst[(by * linesize + bx - 1 - z) * 4] - + odst[((by - 1) * linesize + bx - 1 - z) * 4]; + g = odst[((by - 1) * linesize + bx) * 4 + 1] + + odst[(by * linesize + bx - 1 - z) * 4 + 1] - + odst[((by - 1) * linesize + bx - 1 - z) * 4 + 1]; + b = odst[((by - 1) * linesize + bx) * 4 + 2] + + odst[(by * linesize + bx - 1 - z) * 4 + 2] - + odst[((by - 1) * linesize + bx - 1 - z) * 4 + 2]; + clr = ((b & 0xFF) << 16) + ((g & 0xFF) << 8) + (r & 0xFF); + dst[by * linesize + bx] = clr; + bx++; + if (bx >= x * 16 + sx2 || bx >= avctx->width) { + bx = x * 16 + sx1; + by++; + } + } + break; + case 5: + while (run-- > 0) { + if (by < 1 || by >= avctx->height) + return AVERROR_INVALIDDATA; + + if (bx == 0) { + if (by < 2) + return AVERROR_INVALIDDATA; + z = backstep; + } else { + z = 0; + } + + clr = dst[(by - 1) * linesize + bx - 1 - z]; + dst[by * linesize + bx] = clr; + bx++; + if (bx >= x * 16 + sx2 || bx >= avctx->width) { + bx = x * 16 + sx1; + by++; + } + } + break; + } + + if (avctx->bits_per_coded_sample == 16) { + cx1 = (clr & 0x3F00) >> 2; + cx = (clr & 0x3FFFFF) >> 16; + } else { + cx1 = (clr & 0xFC00) >> 4; + cx = (clr & 0xFFFFFF) >> 18; + } + } + } + } + } + + return 0; +} static void init_rangecoder(RangeCoder *rc, GetByteContext *gb) { @@ -779,18 +2294,26 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, type = bytestream2_peek_byte(gb); if (type == 2) { + s->version = 1; s->get_freq = get_freq0; s->decode = decode0; frame->key_frame = 1; ret = decompress_i(avctx, (uint32_t *)s->current_frame->data[0], s->current_frame->linesize[0] / 4); } else if (type == 18) { + s->version = 2; s->get_freq = get_freq; s->decode = decode; frame->key_frame = 1; ret = decompress_i(avctx, (uint32_t *)s->current_frame->data[0], s->current_frame->linesize[0] / 4); - } else if (type == 17) { + } else if (type == 34) { + frame->key_frame = 1; + s->version = 3; + ret = decompress_i3(avctx, (uint32_t *)s->current_frame->data[0], + s->current_frame->linesize[0] / 4); + ret = 0; + } else if (type == 17 || type == 33) { uint32_t clr, *dst = (uint32_t *)s->current_frame->data[0]; int x, y; @@ -820,10 +2343,16 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame, if (ret < 0) return ret; - ret = decompress_p(avctx, (uint32_t *)s->current_frame->data[0], - s->current_frame->linesize[0] / 4, - (uint32_t *)s->last_frame->data[0], - s->last_frame->linesize[0] / 4); + if (s->version == 1 || s->version == 2) + ret = decompress_p(avctx, (uint32_t *)s->current_frame->data[0], + s->current_frame->linesize[0] / 4, + (uint32_t *)s->last_frame->data[0], + s->last_frame->linesize[0] / 4); + else + ret = decompress_p3(avctx, (uint32_t *)s->current_frame->data[0], + s->current_frame->linesize[0] / 4, + (uint32_t *)s->last_frame->data[0], + s->last_frame->linesize[0] / 4); } else { return AVERROR_PATCHWELCOME; } @@ -914,6 +2443,7 @@ static av_cold int decode_close(AVCodecContext *avctx) av_freep(&s->blocks); av_frame_free(&s->last_frame); av_frame_free(&s->current_frame); + reinit_tables3(s); return 0; } -- 2.17.1 _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel