Shoaib Ahmed / Mbed 2 deprecated uzairkhan

Final 350 project

Dependencies:   uzair Camera_LS_Y201 F7_Ethernet LCD_DISCO_F746NG NetworkAPI SDFileSystem mbed

includes/jfdctfst.c@0:791a779d6220, 2017-07-31 (annotated)

Committer:
shoaib_ahmed
Date:
Mon Jul 31 09:16:35 2017 +0000
Revision:
0:791a779d6220
final project;

Who changed what in which revision?

UserRevisionLine numberNew contents of line
shoaib_ahmed 0:791a779d6220 1 /*
shoaib_ahmed 0:791a779d6220 2 * jfdctfst.c
shoaib_ahmed 0:791a779d6220 3 *
shoaib_ahmed 0:791a779d6220 4 * Copyright (C) 1994-1996, Thomas G. Lane.
shoaib_ahmed 0:791a779d6220 5 * Modified 2003-2015 by Guido Vollbeding.
shoaib_ahmed 0:791a779d6220 6 * This file is part of the Independent JPEG Group's software.
shoaib_ahmed 0:791a779d6220 7 * For conditions of distribution and use, see the accompanying README file.
shoaib_ahmed 0:791a779d6220 8 *
shoaib_ahmed 0:791a779d6220 9 * This file contains a fast, not so accurate integer implementation of the
shoaib_ahmed 0:791a779d6220 10 * forward DCT (Discrete Cosine Transform).
shoaib_ahmed 0:791a779d6220 11 *
shoaib_ahmed 0:791a779d6220 12 * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT
shoaib_ahmed 0:791a779d6220 13 * on each column. Direct algorithms are also available, but they are
shoaib_ahmed 0:791a779d6220 14 * much more complex and seem not to be any faster when reduced to code.
shoaib_ahmed 0:791a779d6220 15 *
shoaib_ahmed 0:791a779d6220 16 * This implementation is based on Arai, Agui, and Nakajima's algorithm for
shoaib_ahmed 0:791a779d6220 17 * scaled DCT. Their original paper (Trans. IEICE E-71(11):1095) is in
shoaib_ahmed 0:791a779d6220 18 * Japanese, but the algorithm is described in the Pennebaker & Mitchell
shoaib_ahmed 0:791a779d6220 19 * JPEG textbook (see REFERENCES section in file README). The following code
shoaib_ahmed 0:791a779d6220 20 * is based directly on figure 4-8 in P&M.
shoaib_ahmed 0:791a779d6220 21 * While an 8-point DCT cannot be done in less than 11 multiplies, it is
shoaib_ahmed 0:791a779d6220 22 * possible to arrange the computation so that many of the multiplies are
shoaib_ahmed 0:791a779d6220 23 * simple scalings of the final outputs. These multiplies can then be
shoaib_ahmed 0:791a779d6220 24 * folded into the multiplications or divisions by the JPEG quantization
shoaib_ahmed 0:791a779d6220 25 * table entries. The AA&N method leaves only 5 multiplies and 29 adds
shoaib_ahmed 0:791a779d6220 26 * to be done in the DCT itself.
shoaib_ahmed 0:791a779d6220 27 * The primary disadvantage of this method is that with fixed-point math,
shoaib_ahmed 0:791a779d6220 28 * accuracy is lost due to imprecise representation of the scaled
shoaib_ahmed 0:791a779d6220 29 * quantization values. The smaller the quantization table entry, the less
shoaib_ahmed 0:791a779d6220 30 * precise the scaled value, so this implementation does worse with high-
shoaib_ahmed 0:791a779d6220 31 * quality-setting files than with low-quality ones.
shoaib_ahmed 0:791a779d6220 32 */
shoaib_ahmed 0:791a779d6220 33
shoaib_ahmed 0:791a779d6220 34 #define JPEG_INTERNALS
shoaib_ahmed 0:791a779d6220 35 #include "jinclude.h"
shoaib_ahmed 0:791a779d6220 36 #include "jpeglib.h"
shoaib_ahmed 0:791a779d6220 37 #include "jdct.h" /* Private declarations for DCT subsystem */
shoaib_ahmed 0:791a779d6220 38
shoaib_ahmed 0:791a779d6220 39 #ifdef DCT_IFAST_SUPPORTED
shoaib_ahmed 0:791a779d6220 40
shoaib_ahmed 0:791a779d6220 41
shoaib_ahmed 0:791a779d6220 42 /*
shoaib_ahmed 0:791a779d6220 43 * This module is specialized to the case DCTSIZE = 8.
shoaib_ahmed 0:791a779d6220 44 */
shoaib_ahmed 0:791a779d6220 45
shoaib_ahmed 0:791a779d6220 46 #if DCTSIZE != 8
shoaib_ahmed 0:791a779d6220 47 Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
shoaib_ahmed 0:791a779d6220 48 #endif
shoaib_ahmed 0:791a779d6220 49
shoaib_ahmed 0:791a779d6220 50
shoaib_ahmed 0:791a779d6220 51 /* Scaling decisions are generally the same as in the LL&M algorithm;
shoaib_ahmed 0:791a779d6220 52 * see jfdctint.c for more details. However, we choose to descale
shoaib_ahmed 0:791a779d6220 53 * (right shift) multiplication products as soon as they are formed,
shoaib_ahmed 0:791a779d6220 54 * rather than carrying additional fractional bits into subsequent additions.
shoaib_ahmed 0:791a779d6220 55 * This compromises accuracy slightly, but it lets us save a few shifts.
shoaib_ahmed 0:791a779d6220 56 * More importantly, 16-bit arithmetic is then adequate (for 8-bit samples)
shoaib_ahmed 0:791a779d6220 57 * everywhere except in the multiplications proper; this saves a good deal
shoaib_ahmed 0:791a779d6220 58 * of work on 16-bit-int machines.
shoaib_ahmed 0:791a779d6220 59 *
shoaib_ahmed 0:791a779d6220 60 * Again to save a few shifts, the intermediate results between pass 1 and
shoaib_ahmed 0:791a779d6220 61 * pass 2 are not upscaled, but are represented only to integral precision.
shoaib_ahmed 0:791a779d6220 62 *
shoaib_ahmed 0:791a779d6220 63 * A final compromise is to represent the multiplicative constants to only
shoaib_ahmed 0:791a779d6220 64 * 8 fractional bits, rather than 13. This saves some shifting work on some
shoaib_ahmed 0:791a779d6220 65 * machines, and may also reduce the cost of multiplication (since there
shoaib_ahmed 0:791a779d6220 66 * are fewer one-bits in the constants).
shoaib_ahmed 0:791a779d6220 67 */
shoaib_ahmed 0:791a779d6220 68
shoaib_ahmed 0:791a779d6220 69 #define CONST_BITS 8
shoaib_ahmed 0:791a779d6220 70
shoaib_ahmed 0:791a779d6220 71
shoaib_ahmed 0:791a779d6220 72 /* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
shoaib_ahmed 0:791a779d6220 73 * causing a lot of useless floating-point operations at run time.
shoaib_ahmed 0:791a779d6220 74 * To get around this we use the following pre-calculated constants.
shoaib_ahmed 0:791a779d6220 75 * If you change CONST_BITS you may want to add appropriate values.
shoaib_ahmed 0:791a779d6220 76 * (With a reasonable C compiler, you can just rely on the FIX() macro...)
shoaib_ahmed 0:791a779d6220 77 */
shoaib_ahmed 0:791a779d6220 78
shoaib_ahmed 0:791a779d6220 79 #if CONST_BITS == 8
shoaib_ahmed 0:791a779d6220 80 #define FIX_0_382683433 ((INT32) 98) /* FIX(0.382683433) */
shoaib_ahmed 0:791a779d6220 81 #define FIX_0_541196100 ((INT32) 139) /* FIX(0.541196100) */
shoaib_ahmed 0:791a779d6220 82 #define FIX_0_707106781 ((INT32) 181) /* FIX(0.707106781) */
shoaib_ahmed 0:791a779d6220 83 #define FIX_1_306562965 ((INT32) 334) /* FIX(1.306562965) */
shoaib_ahmed 0:791a779d6220 84 #else
shoaib_ahmed 0:791a779d6220 85 #define FIX_0_382683433 FIX(0.382683433)
shoaib_ahmed 0:791a779d6220 86 #define FIX_0_541196100 FIX(0.541196100)
shoaib_ahmed 0:791a779d6220 87 #define FIX_0_707106781 FIX(0.707106781)
shoaib_ahmed 0:791a779d6220 88 #define FIX_1_306562965 FIX(1.306562965)
shoaib_ahmed 0:791a779d6220 89 #endif
shoaib_ahmed 0:791a779d6220 90
shoaib_ahmed 0:791a779d6220 91
shoaib_ahmed 0:791a779d6220 92 /* We can gain a little more speed, with a further compromise in accuracy,
shoaib_ahmed 0:791a779d6220 93 * by omitting the addition in a descaling shift. This yields an incorrectly
shoaib_ahmed 0:791a779d6220 94 * rounded result half the time...
shoaib_ahmed 0:791a779d6220 95 */
shoaib_ahmed 0:791a779d6220 96
shoaib_ahmed 0:791a779d6220 97 #ifndef USE_ACCURATE_ROUNDING
shoaib_ahmed 0:791a779d6220 98 #undef DESCALE
shoaib_ahmed 0:791a779d6220 99 #define DESCALE(x,n) RIGHT_SHIFT(x, n)
shoaib_ahmed 0:791a779d6220 100 #endif
shoaib_ahmed 0:791a779d6220 101
shoaib_ahmed 0:791a779d6220 102
shoaib_ahmed 0:791a779d6220 103 /* Multiply a DCTELEM variable by an INT32 constant, and immediately
shoaib_ahmed 0:791a779d6220 104 * descale to yield a DCTELEM result.
shoaib_ahmed 0:791a779d6220 105 */
shoaib_ahmed 0:791a779d6220 106
shoaib_ahmed 0:791a779d6220 107 #define MULTIPLY(var,const) ((DCTELEM) DESCALE((var) * (const), CONST_BITS))
shoaib_ahmed 0:791a779d6220 108
shoaib_ahmed 0:791a779d6220 109
shoaib_ahmed 0:791a779d6220 110 /*
shoaib_ahmed 0:791a779d6220 111 * Perform the forward DCT on one block of samples.
shoaib_ahmed 0:791a779d6220 112 *
shoaib_ahmed 0:791a779d6220 113 * cK represents cos(K*pi/16).
shoaib_ahmed 0:791a779d6220 114 */
shoaib_ahmed 0:791a779d6220 115
shoaib_ahmed 0:791a779d6220 116 GLOBAL(void)
shoaib_ahmed 0:791a779d6220 117 jpeg_fdct_ifast (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col)
shoaib_ahmed 0:791a779d6220 118 {
shoaib_ahmed 0:791a779d6220 119 DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
shoaib_ahmed 0:791a779d6220 120 DCTELEM tmp10, tmp11, tmp12, tmp13;
shoaib_ahmed 0:791a779d6220 121 DCTELEM z1, z2, z3, z4, z5, z11, z13;
shoaib_ahmed 0:791a779d6220 122 DCTELEM *dataptr;
shoaib_ahmed 0:791a779d6220 123 JSAMPROW elemptr;
shoaib_ahmed 0:791a779d6220 124 int ctr;
shoaib_ahmed 0:791a779d6220 125 SHIFT_TEMPS
shoaib_ahmed 0:791a779d6220 126
shoaib_ahmed 0:791a779d6220 127 /* Pass 1: process rows. */
shoaib_ahmed 0:791a779d6220 128
shoaib_ahmed 0:791a779d6220 129 dataptr = data;
shoaib_ahmed 0:791a779d6220 130 for (ctr = 0; ctr < DCTSIZE; ctr++) {
shoaib_ahmed 0:791a779d6220 131 elemptr = sample_data[ctr] + start_col;
shoaib_ahmed 0:791a779d6220 132
shoaib_ahmed 0:791a779d6220 133 /* Load data into workspace */
shoaib_ahmed 0:791a779d6220 134 tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[7]);
shoaib_ahmed 0:791a779d6220 135 tmp7 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[7]);
shoaib_ahmed 0:791a779d6220 136 tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[6]);
shoaib_ahmed 0:791a779d6220 137 tmp6 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[6]);
shoaib_ahmed 0:791a779d6220 138 tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[5]);
shoaib_ahmed 0:791a779d6220 139 tmp5 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[5]);
shoaib_ahmed 0:791a779d6220 140 tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[4]);
shoaib_ahmed 0:791a779d6220 141 tmp4 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[4]);
shoaib_ahmed 0:791a779d6220 142
shoaib_ahmed 0:791a779d6220 143 /* Even part */
shoaib_ahmed 0:791a779d6220 144
shoaib_ahmed 0:791a779d6220 145 tmp10 = tmp0 + tmp3; /* phase 2 */
shoaib_ahmed 0:791a779d6220 146 tmp13 = tmp0 - tmp3;
shoaib_ahmed 0:791a779d6220 147 tmp11 = tmp1 + tmp2;
shoaib_ahmed 0:791a779d6220 148 tmp12 = tmp1 - tmp2;
shoaib_ahmed 0:791a779d6220 149
shoaib_ahmed 0:791a779d6220 150 /* Apply unsigned->signed conversion. */
shoaib_ahmed 0:791a779d6220 151 dataptr[0] = tmp10 + tmp11 - 8 * CENTERJSAMPLE; /* phase 3 */
shoaib_ahmed 0:791a779d6220 152 dataptr[4] = tmp10 - tmp11;
shoaib_ahmed 0:791a779d6220 153
shoaib_ahmed 0:791a779d6220 154 z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
shoaib_ahmed 0:791a779d6220 155 dataptr[2] = tmp13 + z1; /* phase 5 */
shoaib_ahmed 0:791a779d6220 156 dataptr[6] = tmp13 - z1;
shoaib_ahmed 0:791a779d6220 157
shoaib_ahmed 0:791a779d6220 158 /* Odd part */
shoaib_ahmed 0:791a779d6220 159
shoaib_ahmed 0:791a779d6220 160 tmp10 = tmp4 + tmp5; /* phase 2 */
shoaib_ahmed 0:791a779d6220 161 tmp11 = tmp5 + tmp6;
shoaib_ahmed 0:791a779d6220 162 tmp12 = tmp6 + tmp7;
shoaib_ahmed 0:791a779d6220 163
shoaib_ahmed 0:791a779d6220 164 /* The rotator is modified from fig 4-8 to avoid extra negations. */
shoaib_ahmed 0:791a779d6220 165 z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */
shoaib_ahmed 0:791a779d6220 166 z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */
shoaib_ahmed 0:791a779d6220 167 z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
shoaib_ahmed 0:791a779d6220 168 z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
shoaib_ahmed 0:791a779d6220 169
shoaib_ahmed 0:791a779d6220 170 z11 = tmp7 + z3; /* phase 5 */
shoaib_ahmed 0:791a779d6220 171 z13 = tmp7 - z3;
shoaib_ahmed 0:791a779d6220 172
shoaib_ahmed 0:791a779d6220 173 dataptr[5] = z13 + z2; /* phase 6 */
shoaib_ahmed 0:791a779d6220 174 dataptr[3] = z13 - z2;
shoaib_ahmed 0:791a779d6220 175 dataptr[1] = z11 + z4;
shoaib_ahmed 0:791a779d6220 176 dataptr[7] = z11 - z4;
shoaib_ahmed 0:791a779d6220 177
shoaib_ahmed 0:791a779d6220 178 dataptr += DCTSIZE; /* advance pointer to next row */
shoaib_ahmed 0:791a779d6220 179 }
shoaib_ahmed 0:791a779d6220 180
shoaib_ahmed 0:791a779d6220 181 /* Pass 2: process columns. */
shoaib_ahmed 0:791a779d6220 182
shoaib_ahmed 0:791a779d6220 183 dataptr = data;
shoaib_ahmed 0:791a779d6220 184 for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
shoaib_ahmed 0:791a779d6220 185 tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7];
shoaib_ahmed 0:791a779d6220 186 tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7];
shoaib_ahmed 0:791a779d6220 187 tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6];
shoaib_ahmed 0:791a779d6220 188 tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6];
shoaib_ahmed 0:791a779d6220 189 tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5];
shoaib_ahmed 0:791a779d6220 190 tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
shoaib_ahmed 0:791a779d6220 191 tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
shoaib_ahmed 0:791a779d6220 192 tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
shoaib_ahmed 0:791a779d6220 193
shoaib_ahmed 0:791a779d6220 194 /* Even part */
shoaib_ahmed 0:791a779d6220 195
shoaib_ahmed 0:791a779d6220 196 tmp10 = tmp0 + tmp3; /* phase 2 */
shoaib_ahmed 0:791a779d6220 197 tmp13 = tmp0 - tmp3;
shoaib_ahmed 0:791a779d6220 198 tmp11 = tmp1 + tmp2;
shoaib_ahmed 0:791a779d6220 199 tmp12 = tmp1 - tmp2;
shoaib_ahmed 0:791a779d6220 200
shoaib_ahmed 0:791a779d6220 201 dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */
shoaib_ahmed 0:791a779d6220 202 dataptr[DCTSIZE*4] = tmp10 - tmp11;
shoaib_ahmed 0:791a779d6220 203
shoaib_ahmed 0:791a779d6220 204 z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
shoaib_ahmed 0:791a779d6220 205 dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */
shoaib_ahmed 0:791a779d6220 206 dataptr[DCTSIZE*6] = tmp13 - z1;
shoaib_ahmed 0:791a779d6220 207
shoaib_ahmed 0:791a779d6220 208 /* Odd part */
shoaib_ahmed 0:791a779d6220 209
shoaib_ahmed 0:791a779d6220 210 tmp10 = tmp4 + tmp5; /* phase 2 */
shoaib_ahmed 0:791a779d6220 211 tmp11 = tmp5 + tmp6;
shoaib_ahmed 0:791a779d6220 212 tmp12 = tmp6 + tmp7;
shoaib_ahmed 0:791a779d6220 213
shoaib_ahmed 0:791a779d6220 214 /* The rotator is modified from fig 4-8 to avoid extra negations. */
shoaib_ahmed 0:791a779d6220 215 z5 = MULTIPLY(tmp10 - tmp12, FIX_0_382683433); /* c6 */
shoaib_ahmed 0:791a779d6220 216 z2 = MULTIPLY(tmp10, FIX_0_541196100) + z5; /* c2-c6 */
shoaib_ahmed 0:791a779d6220 217 z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
shoaib_ahmed 0:791a779d6220 218 z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
shoaib_ahmed 0:791a779d6220 219
shoaib_ahmed 0:791a779d6220 220 z11 = tmp7 + z3; /* phase 5 */
shoaib_ahmed 0:791a779d6220 221 z13 = tmp7 - z3;
shoaib_ahmed 0:791a779d6220 222
shoaib_ahmed 0:791a779d6220 223 dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */
shoaib_ahmed 0:791a779d6220 224 dataptr[DCTSIZE*3] = z13 - z2;
shoaib_ahmed 0:791a779d6220 225 dataptr[DCTSIZE*1] = z11 + z4;
shoaib_ahmed 0:791a779d6220 226 dataptr[DCTSIZE*7] = z11 - z4;
shoaib_ahmed 0:791a779d6220 227
shoaib_ahmed 0:791a779d6220 228 dataptr++; /* advance pointer to next column */
shoaib_ahmed 0:791a779d6220 229 }
shoaib_ahmed 0:791a779d6220 230 }
shoaib_ahmed 0:791a779d6220 231
shoaib_ahmed 0:791a779d6220 232 #endif /* DCT_IFAST_SUPPORTED */