Subversion Repositories Kolibri OS

Rev

Blame | Last modification | View Log | RSS feed

  1. /*
  2.  * jidctflt.c
  3.  *
  4.  * Copyright (C) 1994-1998, Thomas G. Lane.
  5.  * This file is part of the Independent JPEG Group's software.
  6.  * For conditions of distribution and use, see the accompanying README file.
  7.  *
  8.  * This file contains a floating-point implementation of the
  9.  * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
  10.  * must also perform dequantization of the input coefficients.
  11.  *
  12.  * This implementation should be more accurate than either of the integer
  13.  * IDCT implementations.  However, it may not give the same results on all
  14.  * machines because of differences in roundoff behavior.  Speed will depend
  15.  * on the hardware's floating point capacity.
  16.  *
  17.  * A 2-D IDCT can be done by 1-D IDCT on each column followed by 1-D IDCT
  18.  * on each row (or vice versa, but it's more convenient to emit a row at
  19.  * a time).  Direct algorithms are also available, but they are much more
  20.  * complex and seem not to be any faster when reduced to code.
  21.  *
  22.  * This implementation is based on Arai, Agui, and Nakajima's algorithm for
  23.  * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
  24.  * Japanese, but the algorithm is described in the Pennebaker & Mitchell
  25.  * JPEG textbook (see REFERENCES section in file README).  The following code
  26.  * is based directly on figure 4-8 in P&M.
  27.  * While an 8-point DCT cannot be done in less than 11 multiplies, it is
  28.  * possible to arrange the computation so that many of the multiplies are
  29.  * simple scalings of the final outputs.  These multiplies can then be
  30.  * folded into the multiplications or divisions by the JPEG quantization
  31.  * table entries.  The AA&N method leaves only 5 multiplies and 29 adds
  32.  * to be done in the DCT itself.
  33.  * The primary disadvantage of this method is that with a fixed-point
  34.  * implementation, accuracy is lost due to imprecise representation of the
  35.  * scaled quantization values.  However, that problem does not arise if
  36.  * we use floating point arithmetic.
  37.  */
  38.  
  39. #define JPEG_INTERNALS
  40. #include "jinclude.h"
  41. #include "jpeglib.h"
  42. #include "jdct.h"               /* Private declarations for DCT subsystem */
  43.  
  44. #ifdef DCT_FLOAT_SUPPORTED
  45.  
  46.  
  47. /*
  48.  * This module is specialized to the case DCTSIZE = 8.
  49.  */
  50.  
  51. #if DCTSIZE != 8
  52.   Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
  53. #endif
  54.  
  55.  
  56. /* Dequantize a coefficient by multiplying it by the multiplier-table
  57.  * entry; produce a float result.
  58.  */
  59.  
  60. #define DEQUANTIZE(coef,quantval)  (((FAST_FLOAT) (coef)) * (quantval))
  61.  
  62.  
  63. /*
  64.  * Perform dequantization and inverse DCT on one block of coefficients.
  65.  */
  66.  
  67. GLOBAL(void)
  68. jpeg_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr,
  69.                  JCOEFPTR coef_block,
  70.                  JSAMPARRAY output_buf, JDIMENSION output_col)
  71. {
  72.   FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
  73.   FAST_FLOAT tmp10, tmp11, tmp12, tmp13;
  74.   FAST_FLOAT z5, z10, z11, z12, z13;
  75.   JCOEFPTR inptr;
  76.   FLOAT_MULT_TYPE * quantptr;
  77.   FAST_FLOAT * wsptr;
  78.   JSAMPROW outptr;
  79.   JSAMPLE *range_limit = IDCT_range_limit(cinfo);
  80.   int ctr;
  81.   FAST_FLOAT workspace[DCTSIZE2]; /* buffers data between passes */
  82.   SHIFT_TEMPS
  83.  
  84.   /* Pass 1: process columns from input, store into work array. */
  85.  
  86.   inptr = coef_block;
  87.   quantptr = (FLOAT_MULT_TYPE *) compptr->dct_table;
  88.   wsptr = workspace;
  89.   for (ctr = DCTSIZE; ctr > 0; ctr--) {
  90.     /* Due to quantization, we will usually find that many of the input
  91.      * coefficients are zero, especially the AC terms.  We can exploit this
  92.      * by short-circuiting the IDCT calculation for any column in which all
  93.      * the AC terms are zero.  In that case each output is equal to the
  94.      * DC coefficient (with scale factor as needed).
  95.      * With typical images and quantization tables, half or more of the
  96.      * column DCT calculations can be simplified this way.
  97.      */
  98.    
  99.     if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
  100.         inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
  101.         inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
  102.         inptr[DCTSIZE*7] == 0) {
  103.       /* AC terms all zero */
  104.       FAST_FLOAT dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
  105.      
  106.       wsptr[DCTSIZE*0] = dcval;
  107.       wsptr[DCTSIZE*1] = dcval;
  108.       wsptr[DCTSIZE*2] = dcval;
  109.       wsptr[DCTSIZE*3] = dcval;
  110.       wsptr[DCTSIZE*4] = dcval;
  111.       wsptr[DCTSIZE*5] = dcval;
  112.       wsptr[DCTSIZE*6] = dcval;
  113.       wsptr[DCTSIZE*7] = dcval;
  114.      
  115.       inptr++;                  /* advance pointers to next column */
  116.       quantptr++;
  117.       wsptr++;
  118.       continue;
  119.     }
  120.    
  121.     /* Even part */
  122.  
  123.     tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
  124.     tmp1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
  125.     tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
  126.     tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
  127.  
  128.     tmp10 = tmp0 + tmp2;        /* phase 3 */
  129.     tmp11 = tmp0 - tmp2;
  130.  
  131.     tmp13 = tmp1 + tmp3;        /* phases 5-3 */
  132.     tmp12 = (tmp1 - tmp3) * ((FAST_FLOAT) 1.414213562) - tmp13; /* 2*c4 */
  133.  
  134.     tmp0 = tmp10 + tmp13;       /* phase 2 */
  135.     tmp3 = tmp10 - tmp13;
  136.     tmp1 = tmp11 + tmp12;
  137.     tmp2 = tmp11 - tmp12;
  138.    
  139.     /* Odd part */
  140.  
  141.     tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
  142.     tmp5 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
  143.     tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
  144.     tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
  145.  
  146.     z13 = tmp6 + tmp5;          /* phase 6 */
  147.     z10 = tmp6 - tmp5;
  148.     z11 = tmp4 + tmp7;
  149.     z12 = tmp4 - tmp7;
  150.  
  151.     tmp7 = z11 + z13;           /* phase 5 */
  152.     tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562); /* 2*c4 */
  153.  
  154.     z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */
  155.     tmp10 = ((FAST_FLOAT) 1.082392200) * z12 - z5; /* 2*(c2-c6) */
  156.     tmp12 = ((FAST_FLOAT) -2.613125930) * z10 + z5; /* -2*(c2+c6) */
  157.  
  158.     tmp6 = tmp12 - tmp7;        /* phase 2 */
  159.     tmp5 = tmp11 - tmp6;
  160.     tmp4 = tmp10 + tmp5;
  161.  
  162.     wsptr[DCTSIZE*0] = tmp0 + tmp7;
  163.     wsptr[DCTSIZE*7] = tmp0 - tmp7;
  164.     wsptr[DCTSIZE*1] = tmp1 + tmp6;
  165.     wsptr[DCTSIZE*6] = tmp1 - tmp6;
  166.     wsptr[DCTSIZE*2] = tmp2 + tmp5;
  167.     wsptr[DCTSIZE*5] = tmp2 - tmp5;
  168.     wsptr[DCTSIZE*4] = tmp3 + tmp4;
  169.     wsptr[DCTSIZE*3] = tmp3 - tmp4;
  170.  
  171.     inptr++;                    /* advance pointers to next column */
  172.     quantptr++;
  173.     wsptr++;
  174.   }
  175.  
  176.   /* Pass 2: process rows from work array, store into output array. */
  177.   /* Note that we must descale the results by a factor of 8 == 2**3. */
  178.  
  179.   wsptr = workspace;
  180.   for (ctr = 0; ctr < DCTSIZE; ctr++) {
  181.     outptr = output_buf[ctr] + output_col;
  182.     /* Rows of zeroes can be exploited in the same way as we did with columns.
  183.      * However, the column calculation has created many nonzero AC terms, so
  184.      * the simplification applies less often (typically 5% to 10% of the time).
  185.      * And testing floats for zero is relatively expensive, so we don't bother.
  186.      */
  187.    
  188.     /* Even part */
  189.  
  190.     tmp10 = wsptr[0] + wsptr[4];
  191.     tmp11 = wsptr[0] - wsptr[4];
  192.  
  193.     tmp13 = wsptr[2] + wsptr[6];
  194.     tmp12 = (wsptr[2] - wsptr[6]) * ((FAST_FLOAT) 1.414213562) - tmp13;
  195.  
  196.     tmp0 = tmp10 + tmp13;
  197.     tmp3 = tmp10 - tmp13;
  198.     tmp1 = tmp11 + tmp12;
  199.     tmp2 = tmp11 - tmp12;
  200.  
  201.     /* Odd part */
  202.  
  203.     z13 = wsptr[5] + wsptr[3];
  204.     z10 = wsptr[5] - wsptr[3];
  205.     z11 = wsptr[1] + wsptr[7];
  206.     z12 = wsptr[1] - wsptr[7];
  207.  
  208.     tmp7 = z11 + z13;
  209.     tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562);
  210.  
  211.     z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */
  212.     tmp10 = ((FAST_FLOAT) 1.082392200) * z12 - z5; /* 2*(c2-c6) */
  213.     tmp12 = ((FAST_FLOAT) -2.613125930) * z10 + z5; /* -2*(c2+c6) */
  214.  
  215.     tmp6 = tmp12 - tmp7;
  216.     tmp5 = tmp11 - tmp6;
  217.     tmp4 = tmp10 + tmp5;
  218.  
  219.     /* Final output stage: scale down by a factor of 8 and range-limit */
  220.  
  221.     outptr[0] = range_limit[(int) DESCALE((INT32) (tmp0 + tmp7), 3)
  222.                             & RANGE_MASK];
  223.     outptr[7] = range_limit[(int) DESCALE((INT32) (tmp0 - tmp7), 3)
  224.                             & RANGE_MASK];
  225.     outptr[1] = range_limit[(int) DESCALE((INT32) (tmp1 + tmp6), 3)
  226.                             & RANGE_MASK];
  227.     outptr[6] = range_limit[(int) DESCALE((INT32) (tmp1 - tmp6), 3)
  228.                             & RANGE_MASK];
  229.     outptr[2] = range_limit[(int) DESCALE((INT32) (tmp2 + tmp5), 3)
  230.                             & RANGE_MASK];
  231.     outptr[5] = range_limit[(int) DESCALE((INT32) (tmp2 - tmp5), 3)
  232.                             & RANGE_MASK];
  233.     outptr[4] = range_limit[(int) DESCALE((INT32) (tmp3 + tmp4), 3)
  234.                             & RANGE_MASK];
  235.     outptr[3] = range_limit[(int) DESCALE((INT32) (tmp3 - tmp4), 3)
  236.                             & RANGE_MASK];
  237.    
  238.     wsptr += DCTSIZE;           /* advance pointer to next row */
  239.   }
  240. }
  241.  
  242. #endif /* DCT_FLOAT_SUPPORTED */
  243.