home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Mega CD-ROM 1
/
megacd_rom_1.zip
/
megacd_rom_1
/
MAGAZINE
/
DDJMAG
/
DDJ9102.ZIP
/
NELSON.ZIP
/
MODEL-1.C
< prev
next >
Wrap
C/C++ Source or Header
|
1990-06-02
|
3KB
|
107 lines
/*
* Listing 9 -- model-1.c
*
* This is the modeling module for an order 0 fixed context
* data compression program. This is a relatively simple model.
* The totals for all of the symbols are stored in an array accessed
* under the name "totals". This array has valid indices from -1
* to 256. The reason for having a -1 element is because the EOF
* symbols is included in the table, and it has a value of -1.
*
* The total count for all the symbols is stored in totals[256], and
* the low and high counts for symbol c are found in totals[c] and
* totals[c+1]. The major performance problem with this is that
* the update_model() routine on the average will have to increment
* 128 totals, a very high cost operation.
*/
#include <stdio.h>
#include <stdlib.h>
#include "coder.h"
#include "model.h"
/*
* In order to create an array with indices -1 through 256, I have
* to do this funny declaration. totals[-1] == storage[0].
*/
short int storage[ 258 ];
short int *totals = storage + 1;
/*
* When the model is first started up, each symbols has a count of
* 1, which means a low value of c+1, and a high value of c+2.
*/
void initialize_model()
{
short int i;
for ( i = -1 ; i <= 256 ; i++ )
totals[ i ] = i + 1;
}
/*
* Updating the model means incrementing every single count from
* the high value for the symbol on up to the total. Then, there
* is a complication. If the cumulative total has gone up to
* the maximum value, we need to rescale. Fortunately, the rescale
* operation is relatively rare.
*/
void update_model( int symbol )
{
int i;
for ( symbol++ ; symbol <= 256; symbol++ )
totals[ symbol ]++;
if ( totals[ 256 ] == MAXIMUM_SCALE )
{
for ( i = 0 ; i <= 256 ; i++ )
{
totals[ i ] /= 2;
if ( totals[ i ] <= totals[ i-1 ] )
totals[ i ] = totals[ i-1 ] + 1;
}
}
}
/*
* Finding the low count, high count, and scale for a symbol
* is really easy, because of the way the totals are stored.
* This is the one redeeming feature of the data structure used
* in this implementation. Note that this routine returns an
* int, but it is not used in this routine. The return value
* from convert_int_to_symbol is used in model-2.c.
*/
int convert_int_to_symbol( int c, SYMBOL *s )
{
s->scale = totals[ 256 ];
s->low_count = totals[ c ];
s->high_count = totals[ c+1 ];
return( 0 );
}
/*
* Getting the scale for the current context is easy.
*/
void get_symbol_scale( SYMBOL *s )
{
s->scale = totals[ 256 ];
}
/*
* During decompression, we have to search through the table until
* we find the symbol that straddles the "count" parameter. When
* it is found, it is returned. The reason for also setting the
* high count and low count is so that symbol can be properly removed
* from the arithmetic coded input.
*/
int convert_symbol_to_int( int count, SYMBOL *s)
{
int c;
for ( c = 255; count < totals[ c ] ; c-- )
;
s->high_count = totals[ c+1 ];
s->low_count = totals[ c ];
return( c );
}