1 | // jpge.cpp - C++ class for JPEG compression. Richard Geldreich <richgel99@gmail.com> |
2 | // Supports grayscale, H1V1, H2V1, and H2V2 chroma subsampling factors, one or two pass Huffman table optimization, libjpeg-style quality 1-100 quality factors. |
3 | // Also supports using luma quantization tables for chroma. |
4 | // |
5 | // Released under two licenses. You are free to choose which license you want: |
6 | // License 1: |
7 | // Public Domain |
8 | // |
9 | // License 2: |
10 | // Licensed under the Apache License, Version 2.0 (the "License"); |
11 | // you may not use this file except in compliance with the License. |
12 | // You may obtain a copy of the License at |
13 | // |
14 | // http://www.apache.org/licenses/LICENSE-2.0 |
15 | // |
16 | // Unless required by applicable law or agreed to in writing, software |
17 | // distributed under the License is distributed on an "AS IS" BASIS, |
18 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
19 | // See the License for the specific language governing permissions and |
20 | // limitations under the License. |
21 | // |
22 | // v1.01, Dec. 18, 2010 - Initial release |
23 | // v1.02, Apr. 6, 2011 - Removed 2x2 ordered dither in H2V1 chroma subsampling method load_block_16_8_8(). (The rounding factor was 2, when it should have been 1. Either way, it wasn't helping.) |
24 | // v1.03, Apr. 16, 2011 - Added support for optimized Huffman code tables, optimized dynamic memory allocation down to only 1 alloc. |
25 | // Also from Alex Evans: Added RGBA support, linear memory allocator (no longer needed in v1.03). |
26 | // v1.04, May. 19, 2012: Forgot to set m_pFile ptr to NULL in cfile_stream::close(). Thanks to Owen Kaluza for reporting this bug. |
27 | // Code tweaks to fix VS2008 static code analysis warnings (all looked harmless). |
28 | // Code review revealed method load_block_16_8_8() (used for the non-default H2V1 sampling mode to downsample chroma) somehow didn't get the rounding factor fix from v1.02. |
29 | // v1.05, March 25, 2020: Added Apache 2.0 alternate license |
30 | |
31 | #include "jpge.h" |
32 | |
33 | #include <stdlib.h> |
34 | #include <string.h> |
35 | |
36 | #define JPGE_MAX(a,b) (((a)>(b))?(a):(b)) |
37 | #define JPGE_MIN(a,b) (((a)<(b))?(a):(b)) |
38 | |
39 | namespace jpge { |
40 | |
41 | static inline void* jpge_malloc(size_t nSize) { return malloc(nSize); } |
42 | static inline void jpge_free(void* p) { free(p); } |
43 | |
44 | // Various JPEG enums and tables. |
45 | enum { M_SOF0 = 0xC0, M_DHT = 0xC4, M_SOI = 0xD8, M_EOI = 0xD9, M_SOS = 0xDA, M_DQT = 0xDB, M_APP0 = 0xE0 }; |
46 | enum { DC_LUM_CODES = 12, AC_LUM_CODES = 256, DC_CHROMA_CODES = 12, AC_CHROMA_CODES = 256, MAX_HUFF_SYMBOLS = 257, MAX_HUFF_CODESIZE = 32 }; |
47 | |
48 | static uint8 s_zag[64] = { 0,1,8,16,9,2,3,10,17,24,32,25,18,11,4,5,12,19,26,33,40,48,41,34,27,20,13,6,7,14,21,28,35,42,49,56,57,50,43,36,29,22,15,23,30,37,44,51,58,59,52,45,38,31,39,46,53,60,61,54,47,55,62,63 }; |
49 | static int16 s_std_lum_quant[64] = { 16,11,12,14,12,10,16,14,13,14,18,17,16,19,24,40,26,24,22,22,24,49,35,37,29,40,58,51,61,60,57,51,56,55,64,72,92,78,64,68,87,69,55,56,80,109,81,87,95,98,103,104,103,62,77,113,121,112,100,120,92,101,103,99 }; |
50 | static int16 s_std_croma_quant[64] = { 17,18,18,24,21,24,47,26,26,47,99,66,56,66,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99,99 }; |
51 | |
52 | // Table from http://www.imagemagick.org/discourse-server/viewtopic.php?f=22&t=20333&p=98008#p98008 |
53 | // This is mozjpeg's default table, in zag order. |
54 | static int16 s_alt_quant[64] = { 16,16,16,16,17,16,18,20,20,18,25,27,24,27,25,37,34,31,31,34,37,56,40,43,40,43,40,56,85,53,62,53,53,62,53,85,75,91,74,69,74,91,75,135,106,94,94,106,135,156,131,124,131,156,189,169,169,189,238,226,238,311,311,418 }; |
55 | |
56 | static uint8 s_dc_lum_bits[17] = { 0,0,1,5,1,1,1,1,1,1,0,0,0,0,0,0,0 }; |
57 | static uint8 s_dc_lum_val[DC_LUM_CODES] = { 0,1,2,3,4,5,6,7,8,9,10,11 }; |
58 | static uint8 s_ac_lum_bits[17] = { 0,0,2,1,3,3,2,4,3,5,5,4,4,0,0,1,0x7d }; |
59 | static uint8 s_ac_lum_val[AC_LUM_CODES] = |
60 | { |
61 | 0x01,0x02,0x03,0x00,0x04,0x11,0x05,0x12,0x21,0x31,0x41,0x06,0x13,0x51,0x61,0x07,0x22,0x71,0x14,0x32,0x81,0x91,0xa1,0x08,0x23,0x42,0xb1,0xc1,0x15,0x52,0xd1,0xf0, |
62 | 0x24,0x33,0x62,0x72,0x82,0x09,0x0a,0x16,0x17,0x18,0x19,0x1a,0x25,0x26,0x27,0x28,0x29,0x2a,0x34,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48,0x49, |
63 | 0x4a,0x53,0x54,0x55,0x56,0x57,0x58,0x59,0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x83,0x84,0x85,0x86,0x87,0x88,0x89, |
64 | 0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4,0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xc2,0xc3,0xc4,0xc5, |
65 | 0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xe1,0xe2,0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf1,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8, |
66 | 0xf9,0xfa |
67 | }; |
68 | static uint8 s_dc_chroma_bits[17] = { 0,0,3,1,1,1,1,1,1,1,1,1,0,0,0,0,0 }; |
69 | static uint8 s_dc_chroma_val[DC_CHROMA_CODES] = { 0,1,2,3,4,5,6,7,8,9,10,11 }; |
70 | static uint8 s_ac_chroma_bits[17] = { 0,0,2,1,2,4,4,3,4,7,5,4,4,0,1,2,0x77 }; |
71 | static uint8 s_ac_chroma_val[AC_CHROMA_CODES] = |
72 | { |
73 | 0x00,0x01,0x02,0x03,0x11,0x04,0x05,0x21,0x31,0x06,0x12,0x41,0x51,0x07,0x61,0x71,0x13,0x22,0x32,0x81,0x08,0x14,0x42,0x91,0xa1,0xb1,0xc1,0x09,0x23,0x33,0x52,0xf0, |
74 | 0x15,0x62,0x72,0xd1,0x0a,0x16,0x24,0x34,0xe1,0x25,0xf1,0x17,0x18,0x19,0x1a,0x26,0x27,0x28,0x29,0x2a,0x35,0x36,0x37,0x38,0x39,0x3a,0x43,0x44,0x45,0x46,0x47,0x48, |
75 | 0x49,0x4a,0x53,0x54,0x55,0x56,0x57,0x58,0x59,0x5a,0x63,0x64,0x65,0x66,0x67,0x68,0x69,0x6a,0x73,0x74,0x75,0x76,0x77,0x78,0x79,0x7a,0x82,0x83,0x84,0x85,0x86,0x87, |
76 | 0x88,0x89,0x8a,0x92,0x93,0x94,0x95,0x96,0x97,0x98,0x99,0x9a,0xa2,0xa3,0xa4,0xa5,0xa6,0xa7,0xa8,0xa9,0xaa,0xb2,0xb3,0xb4,0xb5,0xb6,0xb7,0xb8,0xb9,0xba,0xc2,0xc3, |
77 | 0xc4,0xc5,0xc6,0xc7,0xc8,0xc9,0xca,0xd2,0xd3,0xd4,0xd5,0xd6,0xd7,0xd8,0xd9,0xda,0xe2,0xe3,0xe4,0xe5,0xe6,0xe7,0xe8,0xe9,0xea,0xf2,0xf3,0xf4,0xf5,0xf6,0xf7,0xf8, |
78 | 0xf9,0xfa |
79 | }; |
80 | |
81 | // Low-level helper functions. |
82 | template <class T> inline void clear_obj(T& obj) { memset(&obj, 0, sizeof(obj)); } |
83 | |
84 | const int YR = 19595, YG = 38470, YB = 7471, CB_R = -11059, CB_G = -21709, CB_B = 32768, CR_R = 32768, CR_G = -27439, CR_B = -5329; |
85 | static inline uint8 clamp(int i) { if (static_cast<uint>(i) > 255U) { if (i < 0) i = 0; else if (i > 255) i = 255; } return static_cast<uint8>(i); } |
86 | |
87 | static inline int left_shifti(int val, uint32 bits) |
88 | { |
89 | return static_cast<int>(static_cast<uint32>(val) << bits); |
90 | } |
91 | |
92 | static void RGB_to_YCC(uint8* pDst, const uint8* pSrc, int num_pixels) |
93 | { |
94 | for (; num_pixels; pDst += 3, pSrc += 3, num_pixels--) |
95 | { |
96 | const int r = pSrc[0], g = pSrc[1], b = pSrc[2]; |
97 | pDst[0] = static_cast<uint8>((r * YR + g * YG + b * YB + 32768) >> 16); |
98 | pDst[1] = clamp(128 + ((r * CB_R + g * CB_G + b * CB_B + 32768) >> 16)); |
99 | pDst[2] = clamp(128 + ((r * CR_R + g * CR_G + b * CR_B + 32768) >> 16)); |
100 | } |
101 | } |
102 | |
103 | static void RGB_to_Y(uint8* pDst, const uint8* pSrc, int num_pixels) |
104 | { |
105 | for (; num_pixels; pDst++, pSrc += 3, num_pixels--) |
106 | pDst[0] = static_cast<uint8>((pSrc[0] * YR + pSrc[1] * YG + pSrc[2] * YB + 32768) >> 16); |
107 | } |
108 | |
109 | static void RGBA_to_YCC(uint8* pDst, const uint8* pSrc, int num_pixels) |
110 | { |
111 | for (; num_pixels; pDst += 3, pSrc += 4, num_pixels--) |
112 | { |
113 | const int r = pSrc[0], g = pSrc[1], b = pSrc[2]; |
114 | pDst[0] = static_cast<uint8>((r * YR + g * YG + b * YB + 32768) >> 16); |
115 | pDst[1] = clamp(128 + ((r * CB_R + g * CB_G + b * CB_B + 32768) >> 16)); |
116 | pDst[2] = clamp(128 + ((r * CR_R + g * CR_G + b * CR_B + 32768) >> 16)); |
117 | } |
118 | } |
119 | |
120 | static void RGBA_to_Y(uint8* pDst, const uint8* pSrc, int num_pixels) |
121 | { |
122 | for (; num_pixels; pDst++, pSrc += 4, num_pixels--) |
123 | pDst[0] = static_cast<uint8>((pSrc[0] * YR + pSrc[1] * YG + pSrc[2] * YB + 32768) >> 16); |
124 | } |
125 | |
126 | static void Y_to_YCC(uint8* pDst, const uint8* pSrc, int num_pixels) |
127 | { |
128 | for (; num_pixels; pDst += 3, pSrc++, num_pixels--) { pDst[0] = pSrc[0]; pDst[1] = 128; pDst[2] = 128; } |
129 | } |
130 | |
131 | // Forward DCT - DCT derived from jfdctint. |
132 | enum { CONST_BITS = 13, ROW_BITS = 2 }; |
133 | #define DCT_DESCALE(x, n) (((x) + (((int32)1) << ((n) - 1))) >> (n)) |
134 | #define DCT_MUL(var, c) (static_cast<int16>(var) * static_cast<int32>(c)) |
135 | #define DCT1D(s0, s1, s2, s3, s4, s5, s6, s7) \ |
136 | int32 t0 = s0 + s7, t7 = s0 - s7, t1 = s1 + s6, t6 = s1 - s6, t2 = s2 + s5, t5 = s2 - s5, t3 = s3 + s4, t4 = s3 - s4; \ |
137 | int32 t10 = t0 + t3, t13 = t0 - t3, t11 = t1 + t2, t12 = t1 - t2; \ |
138 | int32 u1 = DCT_MUL(t12 + t13, 4433); \ |
139 | s2 = u1 + DCT_MUL(t13, 6270); \ |
140 | s6 = u1 + DCT_MUL(t12, -15137); \ |
141 | u1 = t4 + t7; \ |
142 | int32 u2 = t5 + t6, u3 = t4 + t6, u4 = t5 + t7; \ |
143 | int32 z5 = DCT_MUL(u3 + u4, 9633); \ |
144 | t4 = DCT_MUL(t4, 2446); t5 = DCT_MUL(t5, 16819); \ |
145 | t6 = DCT_MUL(t6, 25172); t7 = DCT_MUL(t7, 12299); \ |
146 | u1 = DCT_MUL(u1, -7373); u2 = DCT_MUL(u2, -20995); \ |
147 | u3 = DCT_MUL(u3, -16069); u4 = DCT_MUL(u4, -3196); \ |
148 | u3 += z5; u4 += z5; \ |
149 | s0 = t10 + t11; s1 = t7 + u1 + u4; s3 = t6 + u2 + u3; s4 = t10 - t11; s5 = t5 + u2 + u4; s7 = t4 + u1 + u3; |
150 | |
151 | static void DCT2D(int32* p) |
152 | { |
153 | int32 c, * q = p; |
154 | for (c = 7; c >= 0; c--, q += 8) |
155 | { |
156 | int32 s0 = q[0], s1 = q[1], s2 = q[2], s3 = q[3], s4 = q[4], s5 = q[5], s6 = q[6], s7 = q[7]; |
157 | DCT1D(s0, s1, s2, s3, s4, s5, s6, s7); |
158 | q[0] = left_shifti(s0, ROW_BITS); q[1] = DCT_DESCALE(s1, CONST_BITS - ROW_BITS); q[2] = DCT_DESCALE(s2, CONST_BITS - ROW_BITS); q[3] = DCT_DESCALE(s3, CONST_BITS - ROW_BITS); |
159 | q[4] = left_shifti(s4, ROW_BITS); q[5] = DCT_DESCALE(s5, CONST_BITS - ROW_BITS); q[6] = DCT_DESCALE(s6, CONST_BITS - ROW_BITS); q[7] = DCT_DESCALE(s7, CONST_BITS - ROW_BITS); |
160 | } |
161 | for (q = p, c = 7; c >= 0; c--, q++) |
162 | { |
163 | int32 s0 = q[0 * 8], s1 = q[1 * 8], s2 = q[2 * 8], s3 = q[3 * 8], s4 = q[4 * 8], s5 = q[5 * 8], s6 = q[6 * 8], s7 = q[7 * 8]; |
164 | DCT1D(s0, s1, s2, s3, s4, s5, s6, s7); |
165 | q[0 * 8] = DCT_DESCALE(s0, ROW_BITS + 3); q[1 * 8] = DCT_DESCALE(s1, CONST_BITS + ROW_BITS + 3); q[2 * 8] = DCT_DESCALE(s2, CONST_BITS + ROW_BITS + 3); q[3 * 8] = DCT_DESCALE(s3, CONST_BITS + ROW_BITS + 3); |
166 | q[4 * 8] = DCT_DESCALE(s4, ROW_BITS + 3); q[5 * 8] = DCT_DESCALE(s5, CONST_BITS + ROW_BITS + 3); q[6 * 8] = DCT_DESCALE(s6, CONST_BITS + ROW_BITS + 3); q[7 * 8] = DCT_DESCALE(s7, CONST_BITS + ROW_BITS + 3); |
167 | } |
168 | } |
169 | |
170 | struct sym_freq { uint m_key, m_sym_index; }; |
171 | |
172 | // Radix sorts sym_freq[] array by 32-bit key m_key. Returns ptr to sorted values. |
173 | static inline sym_freq* radix_sort_syms(uint num_syms, sym_freq* pSyms0, sym_freq* pSyms1) |
174 | { |
175 | const uint cMaxPasses = 4; |
176 | uint32 hist[256 * cMaxPasses]; clear_obj(hist); |
177 | for (uint i = 0; i < num_syms; i++) { uint freq = pSyms0[i].m_key; hist[freq & 0xFF]++; hist[256 + ((freq >> 8) & 0xFF)]++; hist[256 * 2 + ((freq >> 16) & 0xFF)]++; hist[256 * 3 + ((freq >> 24) & 0xFF)]++; } |
178 | sym_freq* pCur_syms = pSyms0, * pNew_syms = pSyms1; |
179 | uint total_passes = cMaxPasses; while ((total_passes > 1) && (num_syms == hist[(total_passes - 1) * 256])) total_passes--; |
180 | for (uint pass_shift = 0, pass = 0; pass < total_passes; pass++, pass_shift += 8) |
181 | { |
182 | const uint32* pHist = &hist[pass << 8]; |
183 | uint offsets[256], cur_ofs = 0; |
184 | for (uint i = 0; i < 256; i++) { offsets[i] = cur_ofs; cur_ofs += pHist[i]; } |
185 | for (uint i = 0; i < num_syms; i++) |
186 | pNew_syms[offsets[(pCur_syms[i].m_key >> pass_shift) & 0xFF]++] = pCur_syms[i]; |
187 | sym_freq* t = pCur_syms; pCur_syms = pNew_syms; pNew_syms = t; |
188 | } |
189 | return pCur_syms; |
190 | } |
191 | |
192 | // calculate_minimum_redundancy() originally written by: Alistair Moffat, alistair@cs.mu.oz.au, Jyrki Katajainen, jyrki@diku.dk, November 1996. |
193 | static void calculate_minimum_redundancy(sym_freq* A, int n) |
194 | { |
195 | int root, leaf, next, avbl, used, dpth; |
196 | if (n == 0) return; else if (n == 1) { A[0].m_key = 1; return; } |
197 | A[0].m_key += A[1].m_key; root = 0; leaf = 2; |
198 | for (next = 1; next < n - 1; next++) |
199 | { |
200 | if (leaf >= n || A[root].m_key < A[leaf].m_key) { A[next].m_key = A[root].m_key; A[root++].m_key = next; } |
201 | else A[next].m_key = A[leaf++].m_key; |
202 | if (leaf >= n || (root < next && A[root].m_key < A[leaf].m_key)) { A[next].m_key += A[root].m_key; A[root++].m_key = next; } |
203 | else A[next].m_key += A[leaf++].m_key; |
204 | } |
205 | A[n - 2].m_key = 0; |
206 | for (next = n - 3; next >= 0; next--) A[next].m_key = A[A[next].m_key].m_key + 1; |
207 | avbl = 1; used = dpth = 0; root = n - 2; next = n - 1; |
208 | while (avbl > 0) |
209 | { |
210 | while (root >= 0 && (int)A[root].m_key == dpth) { used++; root--; } |
211 | while (avbl > used) { A[next--].m_key = dpth; avbl--; } |
212 | avbl = 2 * used; dpth++; used = 0; |
213 | } |
214 | } |
215 | |
216 | // Limits canonical Huffman code table's max code size to max_code_size. |
217 | static void huffman_enforce_max_code_size(int* pNum_codes, int code_list_len, int max_code_size) |
218 | { |
219 | if (code_list_len <= 1) return; |
220 | |
221 | for (int i = max_code_size + 1; i <= MAX_HUFF_CODESIZE; i++) pNum_codes[max_code_size] += pNum_codes[i]; |
222 | |
223 | uint32 total = 0; |
224 | for (int i = max_code_size; i > 0; i--) |
225 | total += (((uint32)pNum_codes[i]) << (max_code_size - i)); |
226 | |
227 | while (total != (1UL << max_code_size)) |
228 | { |
229 | pNum_codes[max_code_size]--; |
230 | for (int i = max_code_size - 1; i > 0; i--) |
231 | { |
232 | if (pNum_codes[i]) { pNum_codes[i]--; pNum_codes[i + 1] += 2; break; } |
233 | } |
234 | total--; |
235 | } |
236 | } |
237 | |
238 | // Generates an optimized offman table. |
239 | void jpeg_encoder::optimize_huffman_table(int table_num, int table_len) |
240 | { |
241 | sym_freq syms0[MAX_HUFF_SYMBOLS], syms1[MAX_HUFF_SYMBOLS]; |
242 | syms0[0].m_key = 1; syms0[0].m_sym_index = 0; // dummy symbol, assures that no valid code contains all 1's |
243 | int num_used_syms = 1; |
244 | const uint32* pSym_count = &m_huff_count[table_num][0]; |
245 | for (int i = 0; i < table_len; i++) |
246 | if (pSym_count[i]) { syms0[num_used_syms].m_key = pSym_count[i]; syms0[num_used_syms++].m_sym_index = i + 1; } |
247 | sym_freq* pSyms = radix_sort_syms(num_used_syms, syms0, syms1); |
248 | calculate_minimum_redundancy(pSyms, num_used_syms); |
249 | |
250 | // Count the # of symbols of each code size. |
251 | int num_codes[1 + MAX_HUFF_CODESIZE]; clear_obj(num_codes); |
252 | for (int i = 0; i < num_used_syms; i++) |
253 | num_codes[pSyms[i].m_key]++; |
254 | |
255 | const uint JPGE_CODE_SIZE_LIMIT = 16; // the maximum possible size of a JPEG Huffman code (valid range is [9,16] - 9 vs. 8 because of the dummy symbol) |
256 | huffman_enforce_max_code_size(num_codes, num_used_syms, JPGE_CODE_SIZE_LIMIT); |
257 | |
258 | // Compute m_huff_bits array, which contains the # of symbols per code size. |
259 | clear_obj(m_huff_bits[table_num]); |
260 | for (int i = 1; i <= (int)JPGE_CODE_SIZE_LIMIT; i++) |
261 | m_huff_bits[table_num][i] = static_cast<uint8>(num_codes[i]); |
262 | |
263 | // Remove the dummy symbol added above, which must be in largest bucket. |
264 | for (int i = JPGE_CODE_SIZE_LIMIT; i >= 1; i--) |
265 | { |
266 | if (m_huff_bits[table_num][i]) { m_huff_bits[table_num][i]--; break; } |
267 | } |
268 | |
269 | // Compute the m_huff_val array, which contains the symbol indices sorted by code size (smallest to largest). |
270 | for (int i = num_used_syms - 1; i >= 1; i--) |
271 | m_huff_val[table_num][num_used_syms - 1 - i] = static_cast<uint8>(pSyms[i].m_sym_index - 1); |
272 | } |
273 | |
274 | // JPEG marker generation. |
275 | void jpeg_encoder::emit_byte(uint8 i) |
276 | { |
277 | m_all_stream_writes_succeeded = m_all_stream_writes_succeeded && m_pStream->put_obj(i); |
278 | } |
279 | |
280 | void jpeg_encoder::emit_word(uint i) |
281 | { |
282 | emit_byte(uint8(i >> 8)); emit_byte(uint8(i & 0xFF)); |
283 | } |
284 | |
285 | void jpeg_encoder::emit_marker(int marker) |
286 | { |
287 | emit_byte(uint8(0xFF)); emit_byte(uint8(marker)); |
288 | } |
289 | |
290 | // Emit JFIF marker |
291 | void jpeg_encoder::emit_jfif_app0() |
292 | { |
293 | emit_marker(M_APP0); |
294 | emit_word(2 + 4 + 1 + 2 + 1 + 2 + 2 + 1 + 1); |
295 | emit_byte(0x4A); emit_byte(0x46); emit_byte(0x49); emit_byte(0x46); /* Identifier: ASCII "JFIF" */ |
296 | emit_byte(0); |
297 | emit_byte(1); /* Major version */ |
298 | emit_byte(1); /* Minor version */ |
299 | emit_byte(0); /* Density unit */ |
300 | emit_word(1); |
301 | emit_word(1); |
302 | emit_byte(0); /* No thumbnail image */ |
303 | emit_byte(0); |
304 | } |
305 | |
306 | // Emit quantization tables |
307 | void jpeg_encoder::emit_dqt() |
308 | { |
309 | for (int i = 0; i < ((m_num_components == 3) ? 2 : 1); i++) |
310 | { |
311 | emit_marker(M_DQT); |
312 | emit_word(64 + 1 + 2); |
313 | emit_byte(static_cast<uint8>(i)); |
314 | for (int j = 0; j < 64; j++) |
315 | emit_byte(static_cast<uint8>(m_quantization_tables[i][j])); |
316 | } |
317 | } |
318 | |
319 | // Emit start of frame marker |
320 | void jpeg_encoder::emit_sof() |
321 | { |
322 | emit_marker(M_SOF0); /* baseline */ |
323 | emit_word(3 * m_num_components + 2 + 5 + 1); |
324 | emit_byte(8); /* precision */ |
325 | emit_word(m_image_y); |
326 | emit_word(m_image_x); |
327 | emit_byte(m_num_components); |
328 | for (int i = 0; i < m_num_components; i++) |
329 | { |
330 | emit_byte(static_cast<uint8>(i + 1)); /* component ID */ |
331 | emit_byte((m_comp_h_samp[i] << 4) + m_comp_v_samp[i]); /* h and v sampling */ |
332 | emit_byte(i > 0); /* quant. table num */ |
333 | } |
334 | } |
335 | |
336 | // Emit Huffman table. |
337 | void jpeg_encoder::emit_dht(uint8* bits, uint8* val, int index, bool ac_flag) |
338 | { |
339 | emit_marker(M_DHT); |
340 | |
341 | int length = 0; |
342 | for (int i = 1; i <= 16; i++) |
343 | length += bits[i]; |
344 | |
345 | emit_word(length + 2 + 1 + 16); |
346 | emit_byte(static_cast<uint8>(index + (ac_flag << 4))); |
347 | |
348 | for (int i = 1; i <= 16; i++) |
349 | emit_byte(bits[i]); |
350 | |
351 | for (int i = 0; i < length; i++) |
352 | emit_byte(val[i]); |
353 | } |
354 | |
355 | // Emit all Huffman tables. |
356 | void jpeg_encoder::emit_dhts() |
357 | { |
358 | emit_dht(m_huff_bits[0 + 0], m_huff_val[0 + 0], 0, false); |
359 | emit_dht(m_huff_bits[2 + 0], m_huff_val[2 + 0], 0, true); |
360 | if (m_num_components == 3) |
361 | { |
362 | emit_dht(m_huff_bits[0 + 1], m_huff_val[0 + 1], 1, false); |
363 | emit_dht(m_huff_bits[2 + 1], m_huff_val[2 + 1], 1, true); |
364 | } |
365 | } |
366 | |
367 | // emit start of scan |
368 | void jpeg_encoder::emit_sos() |
369 | { |
370 | emit_marker(M_SOS); |
371 | emit_word(2 * m_num_components + 2 + 1 + 3); |
372 | emit_byte(m_num_components); |
373 | for (int i = 0; i < m_num_components; i++) |
374 | { |
375 | emit_byte(static_cast<uint8>(i + 1)); |
376 | if (i == 0) |
377 | emit_byte((0 << 4) + 0); |
378 | else |
379 | emit_byte((1 << 4) + 1); |
380 | } |
381 | emit_byte(0); /* spectral selection */ |
382 | emit_byte(63); |
383 | emit_byte(0); |
384 | } |
385 | |
386 | // Emit all markers at beginning of image file. |
387 | void jpeg_encoder::emit_markers() |
388 | { |
389 | emit_marker(M_SOI); |
390 | emit_jfif_app0(); |
391 | emit_dqt(); |
392 | emit_sof(); |
393 | emit_dhts(); |
394 | emit_sos(); |
395 | } |
396 | |
397 | // Compute the actual canonical Huffman codes/code sizes given the JPEG huff bits and val arrays. |
398 | void jpeg_encoder::compute_huffman_table(uint* codes, uint8* code_sizes, uint8* bits, uint8* val) |
399 | { |
400 | int i, l, last_p, si; |
401 | uint8 huff_size[257]; |
402 | uint huff_code[257]; |
403 | uint code; |
404 | |
405 | int p = 0; |
406 | for (l = 1; l <= 16; l++) |
407 | for (i = 1; i <= bits[l]; i++) |
408 | huff_size[p++] = (char)l; |
409 | |
410 | huff_size[p] = 0; last_p = p; // write sentinel |
411 | |
412 | code = 0; si = huff_size[0]; p = 0; |
413 | |
414 | while (huff_size[p]) |
415 | { |
416 | while (huff_size[p] == si) |
417 | huff_code[p++] = code++; |
418 | code <<= 1; |
419 | si++; |
420 | } |
421 | |
422 | memset(codes, 0, sizeof(codes[0]) * 256); |
423 | memset(code_sizes, 0, sizeof(code_sizes[0]) * 256); |
424 | for (p = 0; p < last_p; p++) |
425 | { |
426 | codes[val[p]] = huff_code[p]; |
427 | code_sizes[val[p]] = huff_size[p]; |
428 | } |
429 | } |
430 | |
431 | // Quantization table generation. |
432 | void jpeg_encoder::compute_quant_table(int32* pDst, int16* pSrc) |
433 | { |
434 | int32 q; |
435 | if (m_params.m_quality < 50) |
436 | q = 5000 / m_params.m_quality; |
437 | else |
438 | q = 200 - m_params.m_quality * 2; |
439 | for (int i = 0; i < 64; i++) |
440 | { |
441 | int32 j = *pSrc++; j = (j * q + 50L) / 100L; |
442 | *pDst++ = JPGE_MIN(JPGE_MAX(j, 1), 255); |
443 | } |
444 | } |
445 | |
446 | // Higher-level methods. |
447 | void jpeg_encoder::first_pass_init() |
448 | { |
449 | m_bit_buffer = 0; m_bits_in = 0; |
450 | memset(m_last_dc_val, 0, 3 * sizeof(m_last_dc_val[0])); |
451 | m_mcu_y_ofs = 0; |
452 | m_pass_num = 1; |
453 | } |
454 | |
455 | bool jpeg_encoder::second_pass_init() |
456 | { |
457 | compute_huffman_table(&m_huff_codes[0 + 0][0], &m_huff_code_sizes[0 + 0][0], m_huff_bits[0 + 0], m_huff_val[0 + 0]); |
458 | compute_huffman_table(&m_huff_codes[2 + 0][0], &m_huff_code_sizes[2 + 0][0], m_huff_bits[2 + 0], m_huff_val[2 + 0]); |
459 | if (m_num_components > 1) |
460 | { |
461 | compute_huffman_table(&m_huff_codes[0 + 1][0], &m_huff_code_sizes[0 + 1][0], m_huff_bits[0 + 1], m_huff_val[0 + 1]); |
462 | compute_huffman_table(&m_huff_codes[2 + 1][0], &m_huff_code_sizes[2 + 1][0], m_huff_bits[2 + 1], m_huff_val[2 + 1]); |
463 | } |
464 | first_pass_init(); |
465 | emit_markers(); |
466 | m_pass_num = 2; |
467 | return true; |
468 | } |
469 | |
470 | bool jpeg_encoder::jpg_open(int p_x_res, int p_y_res, int src_channels) |
471 | { |
472 | m_num_components = 3; |
473 | switch (m_params.m_subsampling) |
474 | { |
475 | case Y_ONLY: |
476 | { |
477 | m_num_components = 1; |
478 | m_comp_h_samp[0] = 1; m_comp_v_samp[0] = 1; |
479 | m_mcu_x = 8; m_mcu_y = 8; |
480 | break; |
481 | } |
482 | case H1V1: |
483 | { |
484 | m_comp_h_samp[0] = 1; m_comp_v_samp[0] = 1; |
485 | m_comp_h_samp[1] = 1; m_comp_v_samp[1] = 1; |
486 | m_comp_h_samp[2] = 1; m_comp_v_samp[2] = 1; |
487 | m_mcu_x = 8; m_mcu_y = 8; |
488 | break; |
489 | } |
490 | case H2V1: |
491 | { |
492 | m_comp_h_samp[0] = 2; m_comp_v_samp[0] = 1; |
493 | m_comp_h_samp[1] = 1; m_comp_v_samp[1] = 1; |
494 | m_comp_h_samp[2] = 1; m_comp_v_samp[2] = 1; |
495 | m_mcu_x = 16; m_mcu_y = 8; |
496 | break; |
497 | } |
498 | case H2V2: |
499 | { |
500 | m_comp_h_samp[0] = 2; m_comp_v_samp[0] = 2; |
501 | m_comp_h_samp[1] = 1; m_comp_v_samp[1] = 1; |
502 | m_comp_h_samp[2] = 1; m_comp_v_samp[2] = 1; |
503 | m_mcu_x = 16; m_mcu_y = 16; |
504 | } |
505 | } |
506 | |
507 | m_image_x = p_x_res; m_image_y = p_y_res; |
508 | m_image_bpp = src_channels; |
509 | m_image_bpl = m_image_x * src_channels; |
510 | m_image_x_mcu = (m_image_x + m_mcu_x - 1) & (~(m_mcu_x - 1)); |
511 | m_image_y_mcu = (m_image_y + m_mcu_y - 1) & (~(m_mcu_y - 1)); |
512 | m_image_bpl_xlt = m_image_x * m_num_components; |
513 | m_image_bpl_mcu = m_image_x_mcu * m_num_components; |
514 | m_mcus_per_row = m_image_x_mcu / m_mcu_x; |
515 | |
516 | if ((m_mcu_lines[0] = static_cast<uint8*>(jpge_malloc(m_image_bpl_mcu * m_mcu_y))) == NULL) return false; |
517 | for (int i = 1; i < m_mcu_y; i++) |
518 | m_mcu_lines[i] = m_mcu_lines[i - 1] + m_image_bpl_mcu; |
519 | |
520 | if (m_params.m_use_std_tables) |
521 | { |
522 | compute_quant_table(m_quantization_tables[0], s_std_lum_quant); |
523 | compute_quant_table(m_quantization_tables[1], m_params.m_no_chroma_discrim_flag ? s_std_lum_quant : s_std_croma_quant); |
524 | } |
525 | else |
526 | { |
527 | compute_quant_table(m_quantization_tables[0], s_alt_quant); |
528 | memcpy(m_quantization_tables[1], m_quantization_tables[0], sizeof(m_quantization_tables[1])); |
529 | } |
530 | |
531 | m_out_buf_left = JPGE_OUT_BUF_SIZE; |
532 | m_pOut_buf = m_out_buf; |
533 | |
534 | if (m_params.m_two_pass_flag) |
535 | { |
536 | clear_obj(m_huff_count); |
537 | first_pass_init(); |
538 | } |
539 | else |
540 | { |
541 | memcpy(m_huff_bits[0 + 0], s_dc_lum_bits, 17); memcpy(m_huff_val[0 + 0], s_dc_lum_val, DC_LUM_CODES); |
542 | memcpy(m_huff_bits[2 + 0], s_ac_lum_bits, 17); memcpy(m_huff_val[2 + 0], s_ac_lum_val, AC_LUM_CODES); |
543 | memcpy(m_huff_bits[0 + 1], s_dc_chroma_bits, 17); memcpy(m_huff_val[0 + 1], s_dc_chroma_val, DC_CHROMA_CODES); |
544 | memcpy(m_huff_bits[2 + 1], s_ac_chroma_bits, 17); memcpy(m_huff_val[2 + 1], s_ac_chroma_val, AC_CHROMA_CODES); |
545 | if (!second_pass_init()) return false; // in effect, skip over the first pass |
546 | } |
547 | return m_all_stream_writes_succeeded; |
548 | } |
549 | |
550 | void jpeg_encoder::load_block_8_8_grey(int x) |
551 | { |
552 | uint8* pSrc; |
553 | sample_array_t* pDst = m_sample_array; |
554 | x <<= 3; |
555 | for (int i = 0; i < 8; i++, pDst += 8) |
556 | { |
557 | pSrc = m_mcu_lines[i] + x; |
558 | pDst[0] = pSrc[0] - 128; pDst[1] = pSrc[1] - 128; pDst[2] = pSrc[2] - 128; pDst[3] = pSrc[3] - 128; |
559 | pDst[4] = pSrc[4] - 128; pDst[5] = pSrc[5] - 128; pDst[6] = pSrc[6] - 128; pDst[7] = pSrc[7] - 128; |
560 | } |
561 | } |
562 | |
563 | void jpeg_encoder::load_block_8_8(int x, int y, int c) |
564 | { |
565 | uint8* pSrc; |
566 | sample_array_t* pDst = m_sample_array; |
567 | x = (x * (8 * 3)) + c; |
568 | y <<= 3; |
569 | for (int i = 0; i < 8; i++, pDst += 8) |
570 | { |
571 | pSrc = m_mcu_lines[y + i] + x; |
572 | pDst[0] = pSrc[0 * 3] - 128; pDst[1] = pSrc[1 * 3] - 128; pDst[2] = pSrc[2 * 3] - 128; pDst[3] = pSrc[3 * 3] - 128; |
573 | pDst[4] = pSrc[4 * 3] - 128; pDst[5] = pSrc[5 * 3] - 128; pDst[6] = pSrc[6 * 3] - 128; pDst[7] = pSrc[7 * 3] - 128; |
574 | } |
575 | } |
576 | |
577 | void jpeg_encoder::load_block_16_8(int x, int c) |
578 | { |
579 | uint8* pSrc1, * pSrc2; |
580 | sample_array_t* pDst = m_sample_array; |
581 | x = (x * (16 * 3)) + c; |
582 | for (int i = 0; i < 16; i += 2, pDst += 8) |
583 | { |
584 | pSrc1 = m_mcu_lines[i + 0] + x; |
585 | pSrc2 = m_mcu_lines[i + 1] + x; |
586 | pDst[0] = ((pSrc1[0 * 3] + pSrc1[1 * 3] + pSrc2[0 * 3] + pSrc2[1 * 3] + 2) >> 2) - 128; pDst[1] = ((pSrc1[2 * 3] + pSrc1[3 * 3] + pSrc2[2 * 3] + pSrc2[3 * 3] + 2) >> 2) - 128; |
587 | pDst[2] = ((pSrc1[4 * 3] + pSrc1[5 * 3] + pSrc2[4 * 3] + pSrc2[5 * 3] + 2) >> 2) - 128; pDst[3] = ((pSrc1[6 * 3] + pSrc1[7 * 3] + pSrc2[6 * 3] + pSrc2[7 * 3] + 2) >> 2) - 128; |
588 | pDst[4] = ((pSrc1[8 * 3] + pSrc1[9 * 3] + pSrc2[8 * 3] + pSrc2[9 * 3] + 2) >> 2) - 128; pDst[5] = ((pSrc1[10 * 3] + pSrc1[11 * 3] + pSrc2[10 * 3] + pSrc2[11 * 3] + 2) >> 2) - 128; |
589 | pDst[6] = ((pSrc1[12 * 3] + pSrc1[13 * 3] + pSrc2[12 * 3] + pSrc2[13 * 3] + 2) >> 2) - 128; pDst[7] = ((pSrc1[14 * 3] + pSrc1[15 * 3] + pSrc2[14 * 3] + pSrc2[15 * 3] + 2) >> 2) - 128; |
590 | } |
591 | } |
592 | |
593 | void jpeg_encoder::load_block_16_8_8(int x, int c) |
594 | { |
595 | uint8* pSrc1; |
596 | sample_array_t* pDst = m_sample_array; |
597 | x = (x * (16 * 3)) + c; |
598 | for (int i = 0; i < 8; i++, pDst += 8) |
599 | { |
600 | pSrc1 = m_mcu_lines[i + 0] + x; |
601 | pDst[0] = ((pSrc1[0 * 3] + pSrc1[1 * 3] + 1) >> 1) - 128; pDst[1] = ((pSrc1[2 * 3] + pSrc1[3 * 3] + 1) >> 1) - 128; |
602 | pDst[2] = ((pSrc1[4 * 3] + pSrc1[5 * 3] + 1) >> 1) - 128; pDst[3] = ((pSrc1[6 * 3] + pSrc1[7 * 3] + 1) >> 1) - 128; |
603 | pDst[4] = ((pSrc1[8 * 3] + pSrc1[9 * 3] + 1) >> 1) - 128; pDst[5] = ((pSrc1[10 * 3] + pSrc1[11 * 3] + 1) >> 1) - 128; |
604 | pDst[6] = ((pSrc1[12 * 3] + pSrc1[13 * 3] + 1) >> 1) - 128; pDst[7] = ((pSrc1[14 * 3] + pSrc1[15 * 3] + 1) >> 1) - 128; |
605 | } |
606 | } |
607 | |
608 | void jpeg_encoder::load_quantized_coefficients(int component_num) |
609 | { |
610 | int32* q = m_quantization_tables[component_num > 0]; |
611 | int16* pDst = m_coefficient_array; |
612 | for (int i = 0; i < 64; i++) |
613 | { |
614 | sample_array_t j = m_sample_array[s_zag[i]]; |
615 | if (j < 0) |
616 | { |
617 | if ((j = -j + (*q >> 1)) < *q) |
618 | *pDst++ = 0; |
619 | else |
620 | *pDst++ = static_cast<int16>(-(j / *q)); |
621 | } |
622 | else |
623 | { |
624 | if ((j = j + (*q >> 1)) < *q) |
625 | *pDst++ = 0; |
626 | else |
627 | *pDst++ = static_cast<int16>((j / *q)); |
628 | } |
629 | q++; |
630 | } |
631 | } |
632 | |
633 | void jpeg_encoder::flush_output_buffer() |
634 | { |
635 | if (m_out_buf_left != JPGE_OUT_BUF_SIZE) |
636 | m_all_stream_writes_succeeded = m_all_stream_writes_succeeded && m_pStream->put_buf(m_out_buf, JPGE_OUT_BUF_SIZE - m_out_buf_left); |
637 | m_pOut_buf = m_out_buf; |
638 | m_out_buf_left = JPGE_OUT_BUF_SIZE; |
639 | } |
640 | |
641 | void jpeg_encoder::put_bits(uint bits, uint len) |
642 | { |
643 | m_bit_buffer |= ((uint32)bits << (24 - (m_bits_in += len))); |
644 | while (m_bits_in >= 8) |
645 | { |
646 | uint8 c; |
647 | #define JPGE_PUT_BYTE(c) { *m_pOut_buf++ = (c); if (--m_out_buf_left == 0) flush_output_buffer(); } |
648 | JPGE_PUT_BYTE(c = (uint8)((m_bit_buffer >> 16) & 0xFF)); |
649 | if (c == 0xFF) JPGE_PUT_BYTE(0); |
650 | m_bit_buffer <<= 8; |
651 | m_bits_in -= 8; |
652 | } |
653 | } |
654 | |
655 | void jpeg_encoder::code_coefficients_pass_one(int component_num) |
656 | { |
657 | if (component_num >= 3) return; // just to shut up static analysis |
658 | int i, run_len, nbits, temp1; |
659 | int16* src = m_coefficient_array; |
660 | uint32* dc_count = component_num ? m_huff_count[0 + 1] : m_huff_count[0 + 0], * ac_count = component_num ? m_huff_count[2 + 1] : m_huff_count[2 + 0]; |
661 | |
662 | temp1 = src[0] - m_last_dc_val[component_num]; |
663 | m_last_dc_val[component_num] = src[0]; |
664 | if (temp1 < 0) temp1 = -temp1; |
665 | |
666 | nbits = 0; |
667 | while (temp1) |
668 | { |
669 | nbits++; temp1 >>= 1; |
670 | } |
671 | |
672 | dc_count[nbits]++; |
673 | for (run_len = 0, i = 1; i < 64; i++) |
674 | { |
675 | if ((temp1 = m_coefficient_array[i]) == 0) |
676 | run_len++; |
677 | else |
678 | { |
679 | while (run_len >= 16) |
680 | { |
681 | ac_count[0xF0]++; |
682 | run_len -= 16; |
683 | } |
684 | if (temp1 < 0) temp1 = -temp1; |
685 | nbits = 1; |
686 | while (temp1 >>= 1) nbits++; |
687 | ac_count[(run_len << 4) + nbits]++; |
688 | run_len = 0; |
689 | } |
690 | } |
691 | if (run_len) ac_count[0]++; |
692 | } |
693 | |
694 | void jpeg_encoder::code_coefficients_pass_two(int component_num) |
695 | { |
696 | int i, j, run_len, nbits, temp1, temp2; |
697 | int16* pSrc = m_coefficient_array; |
698 | uint* codes[2]; |
699 | uint8* code_sizes[2]; |
700 | |
701 | if (component_num == 0) |
702 | { |
703 | codes[0] = m_huff_codes[0 + 0]; codes[1] = m_huff_codes[2 + 0]; |
704 | code_sizes[0] = m_huff_code_sizes[0 + 0]; code_sizes[1] = m_huff_code_sizes[2 + 0]; |
705 | } |
706 | else |
707 | { |
708 | codes[0] = m_huff_codes[0 + 1]; codes[1] = m_huff_codes[2 + 1]; |
709 | code_sizes[0] = m_huff_code_sizes[0 + 1]; code_sizes[1] = m_huff_code_sizes[2 + 1]; |
710 | } |
711 | |
712 | temp1 = temp2 = pSrc[0] - m_last_dc_val[component_num]; |
713 | m_last_dc_val[component_num] = pSrc[0]; |
714 | |
715 | if (temp1 < 0) |
716 | { |
717 | temp1 = -temp1; temp2--; |
718 | } |
719 | |
720 | nbits = 0; |
721 | while (temp1) |
722 | { |
723 | nbits++; temp1 >>= 1; |
724 | } |
725 | |
726 | put_bits(codes[0][nbits], code_sizes[0][nbits]); |
727 | if (nbits) put_bits(temp2 & ((1 << nbits) - 1), nbits); |
728 | |
729 | for (run_len = 0, i = 1; i < 64; i++) |
730 | { |
731 | if ((temp1 = m_coefficient_array[i]) == 0) |
732 | run_len++; |
733 | else |
734 | { |
735 | while (run_len >= 16) |
736 | { |
737 | put_bits(codes[1][0xF0], code_sizes[1][0xF0]); |
738 | run_len -= 16; |
739 | } |
740 | if ((temp2 = temp1) < 0) |
741 | { |
742 | temp1 = -temp1; |
743 | temp2--; |
744 | } |
745 | nbits = 1; |
746 | while (temp1 >>= 1) |
747 | nbits++; |
748 | j = (run_len << 4) + nbits; |
749 | put_bits(codes[1][j], code_sizes[1][j]); |
750 | put_bits(temp2 & ((1 << nbits) - 1), nbits); |
751 | run_len = 0; |
752 | } |
753 | } |
754 | if (run_len) |
755 | put_bits(codes[1][0], code_sizes[1][0]); |
756 | } |
757 | |
758 | void jpeg_encoder::code_block(int component_num) |
759 | { |
760 | DCT2D(m_sample_array); |
761 | load_quantized_coefficients(component_num); |
762 | if (m_pass_num == 1) |
763 | code_coefficients_pass_one(component_num); |
764 | else |
765 | code_coefficients_pass_two(component_num); |
766 | } |
767 | |
768 | void jpeg_encoder::process_mcu_row() |
769 | { |
770 | if (m_num_components == 1) |
771 | { |
772 | for (int i = 0; i < m_mcus_per_row; i++) |
773 | { |
774 | load_block_8_8_grey(i); code_block(0); |
775 | } |
776 | } |
777 | else if ((m_comp_h_samp[0] == 1) && (m_comp_v_samp[0] == 1)) |
778 | { |
779 | for (int i = 0; i < m_mcus_per_row; i++) |
780 | { |
781 | load_block_8_8(i, 0, 0); code_block(0); load_block_8_8(i, 0, 1); code_block(1); load_block_8_8(i, 0, 2); code_block(2); |
782 | } |
783 | } |
784 | else if ((m_comp_h_samp[0] == 2) && (m_comp_v_samp[0] == 1)) |
785 | { |
786 | for (int i = 0; i < m_mcus_per_row; i++) |
787 | { |
788 | load_block_8_8(i * 2 + 0, 0, 0); code_block(0); load_block_8_8(i * 2 + 1, 0, 0); code_block(0); |
789 | load_block_16_8_8(i, 1); code_block(1); load_block_16_8_8(i, 2); code_block(2); |
790 | } |
791 | } |
792 | else if ((m_comp_h_samp[0] == 2) && (m_comp_v_samp[0] == 2)) |
793 | { |
794 | for (int i = 0; i < m_mcus_per_row; i++) |
795 | { |
796 | load_block_8_8(i * 2 + 0, 0, 0); code_block(0); load_block_8_8(i * 2 + 1, 0, 0); code_block(0); |
797 | load_block_8_8(i * 2 + 0, 1, 0); code_block(0); load_block_8_8(i * 2 + 1, 1, 0); code_block(0); |
798 | load_block_16_8(i, 1); code_block(1); load_block_16_8(i, 2); code_block(2); |
799 | } |
800 | } |
801 | } |
802 | |
803 | bool jpeg_encoder::terminate_pass_one() |
804 | { |
805 | optimize_huffman_table(0 + 0, DC_LUM_CODES); optimize_huffman_table(2 + 0, AC_LUM_CODES); |
806 | if (m_num_components > 1) |
807 | { |
808 | optimize_huffman_table(0 + 1, DC_CHROMA_CODES); optimize_huffman_table(2 + 1, AC_CHROMA_CODES); |
809 | } |
810 | return second_pass_init(); |
811 | } |
812 | |
813 | bool jpeg_encoder::terminate_pass_two() |
814 | { |
815 | put_bits(0x7F, 7); |
816 | flush_output_buffer(); |
817 | emit_marker(M_EOI); |
818 | m_pass_num++; // purposely bump up m_pass_num, for debugging |
819 | return true; |
820 | } |
821 | |
822 | bool jpeg_encoder::process_end_of_image() |
823 | { |
824 | if (m_mcu_y_ofs) |
825 | { |
826 | if (m_mcu_y_ofs < 16) // check here just to shut up static analysis |
827 | { |
828 | for (int i = m_mcu_y_ofs; i < m_mcu_y; i++) |
829 | memcpy(m_mcu_lines[i], m_mcu_lines[m_mcu_y_ofs - 1], m_image_bpl_mcu); |
830 | } |
831 | |
832 | process_mcu_row(); |
833 | } |
834 | |
835 | if (m_pass_num == 1) |
836 | return terminate_pass_one(); |
837 | else |
838 | return terminate_pass_two(); |
839 | } |
840 | |
841 | void jpeg_encoder::load_mcu(const void* pSrc) |
842 | { |
843 | const uint8* Psrc = reinterpret_cast<const uint8*>(pSrc); |
844 | |
845 | uint8* pDst = m_mcu_lines[m_mcu_y_ofs]; // OK to write up to m_image_bpl_xlt bytes to pDst |
846 | |
847 | if (m_num_components == 1) |
848 | { |
849 | if (m_image_bpp == 4) |
850 | RGBA_to_Y(pDst, Psrc, m_image_x); |
851 | else if (m_image_bpp == 3) |
852 | RGB_to_Y(pDst, Psrc, m_image_x); |
853 | else |
854 | memcpy(pDst, Psrc, m_image_x); |
855 | } |
856 | else |
857 | { |
858 | if (m_image_bpp == 4) |
859 | RGBA_to_YCC(pDst, Psrc, m_image_x); |
860 | else if (m_image_bpp == 3) |
861 | RGB_to_YCC(pDst, Psrc, m_image_x); |
862 | else |
863 | Y_to_YCC(pDst, Psrc, m_image_x); |
864 | } |
865 | |
866 | // Possibly duplicate pixels at end of scanline if not a multiple of 8 or 16 |
867 | if (m_num_components == 1) |
868 | memset(m_mcu_lines[m_mcu_y_ofs] + m_image_bpl_xlt, pDst[m_image_bpl_xlt - 1], m_image_x_mcu - m_image_x); |
869 | else |
870 | { |
871 | const uint8 y = pDst[m_image_bpl_xlt - 3 + 0], cb = pDst[m_image_bpl_xlt - 3 + 1], cr = pDst[m_image_bpl_xlt - 3 + 2]; |
872 | uint8* q = m_mcu_lines[m_mcu_y_ofs] + m_image_bpl_xlt; |
873 | for (int i = m_image_x; i < m_image_x_mcu; i++) |
874 | { |
875 | *q++ = y; *q++ = cb; *q++ = cr; |
876 | } |
877 | } |
878 | |
879 | if (++m_mcu_y_ofs == m_mcu_y) |
880 | { |
881 | process_mcu_row(); |
882 | m_mcu_y_ofs = 0; |
883 | } |
884 | } |
885 | |
886 | void jpeg_encoder::clear() |
887 | { |
888 | m_mcu_lines[0] = NULL; |
889 | m_pass_num = 0; |
890 | m_all_stream_writes_succeeded = true; |
891 | } |
892 | |
893 | jpeg_encoder::jpeg_encoder() |
894 | { |
895 | clear(); |
896 | } |
897 | |
898 | jpeg_encoder::~jpeg_encoder() |
899 | { |
900 | deinit(); |
901 | } |
902 | |
903 | bool jpeg_encoder::init(output_stream* pStream, int width, int height, int src_channels, const params& comp_params) |
904 | { |
905 | deinit(); |
906 | if (((!pStream) || (width < 1) || (height < 1)) || ((src_channels != 1) && (src_channels != 3) && (src_channels != 4)) || (!comp_params.check())) return false; |
907 | m_pStream = pStream; |
908 | m_params = comp_params; |
909 | return jpg_open(width, height, src_channels); |
910 | } |
911 | |
912 | void jpeg_encoder::deinit() |
913 | { |
914 | jpge_free(m_mcu_lines[0]); |
915 | clear(); |
916 | } |
917 | |
918 | bool jpeg_encoder::process_scanline(const void* pScanline) |
919 | { |
920 | if ((m_pass_num < 1) || (m_pass_num > 2)) return false; |
921 | if (m_all_stream_writes_succeeded) |
922 | { |
923 | if (!pScanline) |
924 | { |
925 | if (!process_end_of_image()) return false; |
926 | } |
927 | else |
928 | { |
929 | load_mcu(pScanline); |
930 | } |
931 | } |
932 | return m_all_stream_writes_succeeded; |
933 | } |
934 | |
935 | // Higher level wrappers/examples (optional). |
936 | #include <stdio.h> |
937 | |
938 | class cfile_stream : public output_stream |
939 | { |
940 | cfile_stream(const cfile_stream&); |
941 | cfile_stream& operator= (const cfile_stream&); |
942 | |
943 | FILE* m_pFile; |
944 | bool m_bStatus; |
945 | |
946 | public: |
947 | cfile_stream() : m_pFile(NULL), m_bStatus(false) { } |
948 | |
949 | virtual ~cfile_stream() |
950 | { |
951 | close(); |
952 | } |
953 | |
954 | bool open(const char* pFilename) |
955 | { |
956 | close(); |
957 | m_pFile = fopen(pFilename, "wb" ); |
958 | m_bStatus = (m_pFile != NULL); |
959 | return m_bStatus; |
960 | } |
961 | |
962 | bool close() |
963 | { |
964 | if (m_pFile) |
965 | { |
966 | if (fclose(m_pFile) == EOF) |
967 | { |
968 | m_bStatus = false; |
969 | } |
970 | m_pFile = NULL; |
971 | } |
972 | return m_bStatus; |
973 | } |
974 | |
975 | virtual bool put_buf(const void* pBuf, int len) |
976 | { |
977 | m_bStatus = m_bStatus && (fwrite(pBuf, len, 1, m_pFile) == 1); |
978 | return m_bStatus; |
979 | } |
980 | |
981 | uint get_size() const |
982 | { |
983 | return m_pFile ? ftell(m_pFile) : 0; |
984 | } |
985 | }; |
986 | |
987 | // Writes JPEG image to file. |
988 | bool compress_image_to_jpeg_file(const char* pFilename, int width, int height, int num_channels, const uint8* pImage_data, const params& comp_params) |
989 | { |
990 | cfile_stream dst_stream; |
991 | if (!dst_stream.open(pFilename)) |
992 | return false; |
993 | |
994 | jpge::jpeg_encoder dst_image; |
995 | if (!dst_image.init(&dst_stream, width, height, num_channels, comp_params)) |
996 | return false; |
997 | |
998 | for (uint pass_index = 0; pass_index < dst_image.get_total_passes(); pass_index++) |
999 | { |
1000 | for (int i = 0; i < height; i++) |
1001 | { |
1002 | const uint8* pBuf = pImage_data + i * width * num_channels; |
1003 | if (!dst_image.process_scanline(pBuf)) |
1004 | return false; |
1005 | } |
1006 | if (!dst_image.process_scanline(NULL)) |
1007 | return false; |
1008 | } |
1009 | |
1010 | dst_image.deinit(); |
1011 | |
1012 | return dst_stream.close(); |
1013 | } |
1014 | |
1015 | class memory_stream : public output_stream |
1016 | { |
1017 | memory_stream(const memory_stream&); |
1018 | memory_stream& operator= (const memory_stream&); |
1019 | |
1020 | uint8* m_pBuf; |
1021 | uint m_buf_size, m_buf_ofs; |
1022 | |
1023 | public: |
1024 | memory_stream(void* pBuf, uint buf_size) : m_pBuf(static_cast<uint8*>(pBuf)), m_buf_size(buf_size), m_buf_ofs(0) { } |
1025 | |
1026 | virtual ~memory_stream() { } |
1027 | |
1028 | virtual bool put_buf(const void* pBuf, int len) |
1029 | { |
1030 | uint buf_remaining = m_buf_size - m_buf_ofs; |
1031 | if ((uint)len > buf_remaining) |
1032 | return false; |
1033 | memcpy(m_pBuf + m_buf_ofs, pBuf, len); |
1034 | m_buf_ofs += len; |
1035 | return true; |
1036 | } |
1037 | |
1038 | uint get_size() const |
1039 | { |
1040 | return m_buf_ofs; |
1041 | } |
1042 | }; |
1043 | |
1044 | bool compress_image_to_jpeg_file_in_memory(void* pDstBuf, int& buf_size, int width, int height, int num_channels, const uint8* pImage_data, const params& comp_params) |
1045 | { |
1046 | if ((!pDstBuf) || (!buf_size)) |
1047 | return false; |
1048 | |
1049 | memory_stream dst_stream(pDstBuf, buf_size); |
1050 | |
1051 | buf_size = 0; |
1052 | |
1053 | jpge::jpeg_encoder dst_image; |
1054 | if (!dst_image.init(&dst_stream, width, height, num_channels, comp_params)) |
1055 | return false; |
1056 | |
1057 | for (uint pass_index = 0; pass_index < dst_image.get_total_passes(); pass_index++) |
1058 | { |
1059 | for (int i = 0; i < height; i++) |
1060 | { |
1061 | const uint8* pScanline = pImage_data + i * width * num_channels; |
1062 | if (!dst_image.process_scanline(pScanline)) |
1063 | return false; |
1064 | } |
1065 | if (!dst_image.process_scanline(NULL)) |
1066 | return false; |
1067 | } |
1068 | |
1069 | dst_image.deinit(); |
1070 | |
1071 | buf_size = dst_stream.get_size(); |
1072 | return true; |
1073 | } |
1074 | |
1075 | } // namespace jpge |
1076 | |
1077 | |