CentrED/Imaging/JpegLib/imjcdctmgr.pas

514 lines
16 KiB
Plaintext
Raw Permalink Normal View History

2022-05-08 10:47:53 +02:00
unit imjcdctmgr;
{ Original : jcdctmgr.c ; Copyright (C) 1994-1996, Thomas G. Lane. }
{ This file is part of the Independent JPEG Group's software.
For conditions of distribution and use, see the accompanying README file.
This file contains the forward-DCT management logic.
This code selects a particular DCT implementation to be used,
and it performs related housekeeping chores including coefficient
quantization. }
interface
{$I imjconfig.inc}
uses
imjmorecfg,
imjinclude,
imjdeferr,
imjerror,
imjpeglib,
imjdct, { Private declarations for DCT subsystem }
imjfdctint, imjfdctfst, imjfdctflt;
{ Initialize FDCT manager. }
{GLOBAL}
procedure jinit_forward_dct (cinfo : j_compress_ptr);
implementation
{ Private subobject for this module }
type
my_fdct_ptr = ^my_fdct_controller;
my_fdct_controller = record
pub : jpeg_forward_dct; { public fields }
{ Pointer to the DCT routine actually in use }
do_dct : forward_DCT_method_ptr;
{ The actual post-DCT divisors --- not identical to the quant table
entries, because of scaling (especially for an unnormalized DCT).
Each table is given in normal array order. }
divisors : array[0..NUM_QUANT_TBLS-1] of DCTELEM_FIELD_PTR;
{$ifdef DCT_FLOAT_SUPPORTED}
{ Same as above for the floating-point case. }
do_float_dct : float_DCT_method_ptr;
float_divisors : array[0..NUM_QUANT_TBLS-1] of FAST_FLOAT_FIELD_PTR;
{$endif}
end;
{ Initialize for a processing pass.
Verify that all referenced Q-tables are present, and set up
the divisor table for each one.
In the current implementation, DCT of all components is done during
the first pass, even if only some components will be output in the
first scan. Hence all components should be examined here. }
{METHODDEF}
procedure start_pass_fdctmgr (cinfo : j_compress_ptr);
var
fdct : my_fdct_ptr;
ci, qtblno, i : int;
compptr : jpeg_component_info_ptr;
qtbl : JQUANT_TBL_PTR;
dtbl : DCTELEM_FIELD_PTR;
{$ifdef DCT_IFAST_SUPPORTED}
const
CONST_BITS = 14;
aanscales : array[0..DCTSIZE2-1] of INT16 =
({ precomputed values scaled up by 14 bits }
16384, 22725, 21407, 19266, 16384, 12873, 8867, 4520,
22725, 31521, 29692, 26722, 22725, 17855, 12299, 6270,
21407, 29692, 27969, 25172, 21407, 16819, 11585, 5906,
19266, 26722, 25172, 22654, 19266, 15137, 10426, 5315,
16384, 22725, 21407, 19266, 16384, 12873, 8867, 4520,
12873, 17855, 16819, 15137, 12873, 10114, 6967, 3552,
8867, 12299, 11585, 10426, 8867, 6967, 4799, 2446,
4520, 6270, 5906, 5315, 4520, 3552, 2446, 1247);
{SHIFT_TEMPS}
{ Descale and correctly round an INT32 value that's scaled by N bits.
We assume RIGHT_SHIFT rounds towards minus infinity, so adding
the fudge factor is correct for either sign of X. }
function DESCALE(x : INT32; n : int) : INT32;
var
shift_temp : INT32;
begin
shift_temp := x + (INT32(1) shl (n-1));
{$ifdef RIGHT_SHIFT_IS_UNSIGNED}
if shift_temp < 0 then
Descale := (shift_temp shr n) or ((not INT32(0)) shl (32-n))
else
{$endif}
Descale := (shift_temp shr n);
end;
{$endif}
{$ifdef DCT_FLOAT_SUPPORTED}
var
fdtbl : FAST_FLOAT_FIELD_PTR;
row, col : int;
const
aanscalefactor : array[0..DCTSIZE-1] of double =
(1.0, 1.387039845, 1.306562965, 1.175875602,
1.0, 0.785694958, 0.541196100, 0.275899379);
{$endif}
begin
fdct := my_fdct_ptr (cinfo^.fdct);
compptr := jpeg_component_info_ptr(cinfo^.comp_info);
for ci := 0 to pred(cinfo^.num_components) do
begin
qtblno := compptr^.quant_tbl_no;
{ Make sure specified quantization table is present }
if (qtblno < 0) or (qtblno >= NUM_QUANT_TBLS) or
(cinfo^.quant_tbl_ptrs[qtblno] = NIL) then
ERREXIT1(j_common_ptr(cinfo), JERR_NO_QUANT_TABLE, qtblno);
qtbl := cinfo^.quant_tbl_ptrs[qtblno];
{ Compute divisors for this quant table }
{ We may do this more than once for same table, but it's not a big deal }
case (cinfo^.dct_method) of
{$ifdef DCT_ISLOW_SUPPORTED}
JDCT_ISLOW:
begin
{ For LL&M IDCT method, divisors are equal to raw quantization
coefficients multiplied by 8 (to counteract scaling). }
if (fdct^.divisors[qtblno] = NIL) then
begin
fdct^.divisors[qtblno] := DCTELEM_FIELD_PTR(
cinfo^.mem^.alloc_small (j_common_ptr(cinfo), JPOOL_IMAGE,
DCTSIZE2 * SIZEOF(DCTELEM)) );
end;
dtbl := fdct^.divisors[qtblno];
for i := 0 to pred(DCTSIZE2) do
begin
dtbl^[i] := (DCTELEM(qtbl^.quantval[i])) shl 3;
end;
end;
{$endif}
{$ifdef DCT_IFAST_SUPPORTED}
JDCT_IFAST:
begin
{ For AA&N IDCT method, divisors are equal to quantization
coefficients scaled by scalefactor[row]*scalefactor[col], where
scalefactor[0] := 1
scalefactor[k] := cos(k*PI/16) * sqrt(2) for k=1..7
We apply a further scale factor of 8. }
if (fdct^.divisors[qtblno] = NIL) then
begin
fdct^.divisors[qtblno] := DCTELEM_FIELD_PTR(
cinfo^.mem^.alloc_small (j_common_ptr(cinfo), JPOOL_IMAGE,
DCTSIZE2 * SIZEOF(DCTELEM)) );
end;
dtbl := fdct^.divisors[qtblno];
for i := 0 to pred(DCTSIZE2) do
begin
dtbl^[i] := DCTELEM(
{MULTIPLY16V16}
DESCALE( INT32(qtbl^.quantval[i]) * INT32 (aanscales[i]),
CONST_BITS-3) );
end;
end;
{$endif}
{$ifdef DCT_FLOAT_SUPPORTED}
JDCT_FLOAT:
begin
{ For float AA&N IDCT method, divisors are equal to quantization
coefficients scaled by scalefactor[row]*scalefactor[col], where
scalefactor[0] := 1
scalefactor[k] := cos(k*PI/16) * sqrt(2) for k=1..7
We apply a further scale factor of 8.
What's actually stored is 1/divisor so that the inner loop can
use a multiplication rather than a division. }
if (fdct^.float_divisors[qtblno] = NIL) then
begin
fdct^.float_divisors[qtblno] := FAST_FLOAT_FIELD_PTR(
cinfo^.mem^.alloc_small (j_common_ptr(cinfo), JPOOL_IMAGE,
DCTSIZE2 * SIZEOF(FAST_FLOAT)) );
end;
fdtbl := fdct^.float_divisors[qtblno];
i := 0;
for row := 0 to pred(DCTSIZE) do
begin
for col := 0 to pred(DCTSIZE) do
begin
fdtbl^[i] := {FAST_FLOAT}
(1.0 / (( {double}(qtbl^.quantval[i]) *
aanscalefactor[row] * aanscalefactor[col] * 8.0)));
Inc(i);
end;
end;
end;
{$endif}
else
ERREXIT(j_common_ptr(cinfo), JERR_NOT_COMPILED);
end;
Inc(compptr);
end;
end;
{ Perform forward DCT on one or more blocks of a component.
The input samples are taken from the sample_data[] array starting at
position start_row/start_col, and moving to the right for any additional
blocks. The quantized coefficients are returned in coef_blocks[]. }
{METHODDEF}
procedure forward_DCT (cinfo : j_compress_ptr;
compptr : jpeg_component_info_ptr;
sample_data : JSAMPARRAY;
coef_blocks : JBLOCKROW;
start_row : JDIMENSION;
start_col : JDIMENSION;
num_blocks : JDIMENSION);
{ This version is used for integer DCT implementations. }
var
{ This routine is heavily used, so it's worth coding it tightly. }
fdct : my_fdct_ptr;
do_dct : forward_DCT_method_ptr;
divisors : DCTELEM_FIELD_PTR;
workspace : array[0..DCTSIZE2-1] of DCTELEM; { work area for FDCT subroutine }
bi : JDIMENSION;
var
{register} workspaceptr : DCTELEMPTR;
{register} elemptr : JSAMPLE_PTR;
{register} elemr : int;
{$ifndef DCTSIZE_IS_8}
var
{register} elemc : int;
{$endif}
var
{register} temp, qval : DCTELEM;
{register} i : int;
{register} output_ptr : JCOEFPTR;
begin
fdct := my_fdct_ptr (cinfo^.fdct);
do_dct := fdct^.do_dct;
divisors := fdct^.divisors[compptr^.quant_tbl_no];
Inc(JSAMPROW_PTR(sample_data), start_row); { fold in the vertical offset once }
for bi := 0 to pred(num_blocks) do
begin
{ Load data into workspace, applying unsigned->signed conversion }
workspaceptr := @workspace[0];
for elemr := 0 to pred(DCTSIZE) do
begin
elemptr := @sample_data^[elemr]^[start_col];
{$ifdef DCTSIZE_IS_8} { unroll the inner loop }
workspaceptr^ := GETJSAMPLE(elemptr^) - CENTERJSAMPLE;
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := GETJSAMPLE(elemptr^) - CENTERJSAMPLE;
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := GETJSAMPLE(elemptr^) - CENTERJSAMPLE;
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := GETJSAMPLE(elemptr^) - CENTERJSAMPLE;
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := GETJSAMPLE(elemptr^) - CENTERJSAMPLE;
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := GETJSAMPLE(elemptr^) - CENTERJSAMPLE;
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := GETJSAMPLE(elemptr^) - CENTERJSAMPLE;
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := GETJSAMPLE(elemptr^) - CENTERJSAMPLE;
Inc(workspaceptr);
{Inc(elemptr); - Value never used }
{$else}
for elemc := pred(DCTSIZE) downto 0 do
begin
workspaceptr^ := GETJSAMPLE(elemptr^) - CENTERJSAMPLE;
Inc(workspaceptr);
Inc(elemptr);
end;
{$endif}
end;
{ Perform the DCT }
do_dct (workspace);
{ Quantize/descale the coefficients, and store into coef_blocks[] }
output_ptr := JCOEFPTR(@coef_blocks^[bi]);
for i := 0 to pred(DCTSIZE2) do
begin
qval := divisors^[i];
temp := workspace[i];
{ Divide the coefficient value by qval, ensuring proper rounding.
Since C does not specify the direction of rounding for negative
quotients, we have to force the dividend positive for portability.
In most files, at least half of the output values will be zero
(at default quantization settings, more like three-quarters...)
so we should ensure that this case is fast. On many machines,
a comparison is enough cheaper than a divide to make a special test
a win. Since both inputs will be nonnegative, we need only test
for a < b to discover whether a/b is 0.
If your machine's division is fast enough, define FAST_DIVIDE. }
if (temp < 0) then
begin
temp := -temp;
Inc(temp, qval shr 1); { for rounding }
{DIVIDE_BY(temp, qval);}
{$ifdef FAST_DIVIDE}
temp := temp div qval;
{$else}
if (temp >= qval) then
temp := temp div qval
else
temp := 0;
{$endif}
temp := -temp;
end
else
begin
Inc(temp, qval shr 1); { for rounding }
{DIVIDE_BY(temp, qval);}
{$ifdef FAST_DIVIDE}
temp := temp div qval;
{$else}
if (temp >= qval) then
temp := temp div qval
else
temp := 0;
{$endif}
end;
output_ptr^[i] := JCOEF (temp);
end;
Inc(start_col, DCTSIZE);
end;
end;
{$ifdef DCT_FLOAT_SUPPORTED}
{METHODDEF}
procedure forward_DCT_float (cinfo : j_compress_ptr;
compptr : jpeg_component_info_ptr;
sample_data : JSAMPARRAY;
coef_blocks : JBLOCKROW;
start_row : JDIMENSION;
start_col : JDIMENSION;
num_blocks : JDIMENSION);
{ This version is used for floating-point DCT implementations. }
var
{ This routine is heavily used, so it's worth coding it tightly. }
fdct : my_fdct_ptr;
do_dct : float_DCT_method_ptr;
divisors : FAST_FLOAT_FIELD_PTR;
workspace : array[0..DCTSIZE2-1] of FAST_FLOAT; { work area for FDCT subroutine }
bi : JDIMENSION;
var
{register} workspaceptr : FAST_FLOAT_PTR;
{register} elemptr : JSAMPLE_PTR;
{register} elemr : int;
{$ifndef DCTSIZE_IS_8}
var
{register} elemc : int;
{$endif}
var
{register} temp : FAST_FLOAT;
{register} i : int;
{register} output_ptr : JCOEFPTR;
begin
fdct := my_fdct_ptr (cinfo^.fdct);
do_dct := fdct^.do_float_dct;
divisors := fdct^.float_divisors[compptr^.quant_tbl_no];
Inc(JSAMPROW_PTR(sample_data), start_row); { fold in the vertical offset once }
for bi := 0 to pred(num_blocks) do
begin
{ Load data into workspace, applying unsigned->signed conversion }
workspaceptr := @workspace[0];
for elemr := 0 to pred(DCTSIZE) do
begin
elemptr := @(sample_data^[elemr]^[start_col]);
{$ifdef DCTSIZE_IS_8} { unroll the inner loop }
workspaceptr^ := {FAST_FLOAT}(GETJSAMPLE(elemptr^) - CENTERJSAMPLE);
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := {FAST_FLOAT}(GETJSAMPLE(elemptr^) - CENTERJSAMPLE);
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := {FAST_FLOAT}(GETJSAMPLE(elemptr^) - CENTERJSAMPLE);
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := {FAST_FLOAT}(GETJSAMPLE(elemptr^) - CENTERJSAMPLE);
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := {FAST_FLOAT}(GETJSAMPLE(elemptr^) - CENTERJSAMPLE);
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := {FAST_FLOAT}(GETJSAMPLE(elemptr^) - CENTERJSAMPLE);
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := {FAST_FLOAT}(GETJSAMPLE(elemptr^) - CENTERJSAMPLE);
Inc(workspaceptr);
Inc(elemptr);
workspaceptr^ := {FAST_FLOAT}(GETJSAMPLE(elemptr^) - CENTERJSAMPLE);
Inc(workspaceptr);
{Inc(elemptr); - value never used }
{$else}
for elemc := pred(DCTSIZE) downto 0 do
begin
workspaceptr^ := {FAST_FLOAT}(
(GETJSAMPLE(elemptr^) - CENTERJSAMPLE) );
Inc(workspaceptr);
Inc(elemptr);
end;
{$endif}
end;
{ Perform the DCT }
do_dct (workspace);
{ Quantize/descale the coefficients, and store into coef_blocks[] }
output_ptr := JCOEFPTR(@(coef_blocks^[bi]));
for i := 0 to pred(DCTSIZE2) do
begin
{ Apply the quantization and scaling factor }
temp := workspace[i] * divisors^[i];
{ Round to nearest integer.
Since C does not specify the direction of rounding for negative
quotients, we have to force the dividend positive for portability.
The maximum coefficient size is +-16K (for 12-bit data), so this
code should work for either 16-bit or 32-bit ints. }
output_ptr^[i] := JCOEF ( int(Trunc (temp + {FAST_FLOAT}(16384.5))) - 16384);
end;
Inc(start_col, DCTSIZE);
end;
end;
{$endif} { DCT_FLOAT_SUPPORTED }
{ Initialize FDCT manager. }
{GLOBAL}
procedure jinit_forward_dct (cinfo : j_compress_ptr);
var
fdct : my_fdct_ptr;
i : int;
begin
fdct := my_fdct_ptr(
cinfo^.mem^.alloc_small (j_common_ptr(cinfo), JPOOL_IMAGE,
SIZEOF(my_fdct_controller)) );
cinfo^.fdct := jpeg_forward_dct_ptr (fdct);
fdct^.pub.start_pass := start_pass_fdctmgr;
case (cinfo^.dct_method) of
{$ifdef DCT_ISLOW_SUPPORTED}
JDCT_ISLOW:
begin
fdct^.pub.forward_DCT := forward_DCT;
fdct^.do_dct := jpeg_fdct_islow;
end;
{$endif}
{$ifdef DCT_IFAST_SUPPORTED}
JDCT_IFAST:
begin
fdct^.pub.forward_DCT := forward_DCT;
fdct^.do_dct := jpeg_fdct_ifast;
end;
{$endif}
{$ifdef DCT_FLOAT_SUPPORTED}
JDCT_FLOAT:
begin
fdct^.pub.forward_DCT := forward_DCT_float;
fdct^.do_float_dct := jpeg_fdct_float;
end;
{$endif}
else
ERREXIT(j_common_ptr(cinfo), JERR_NOT_COMPILED);
end;
{ Mark divisor tables unallocated }
for i := 0 to pred(NUM_QUANT_TBLS) do
begin
fdct^.divisors[i] := NIL;
{$ifdef DCT_FLOAT_SUPPORTED}
fdct^.float_divisors[i] := NIL;
{$endif}
end;
end;
end.