de.sciss.jump3r.mp3

Class Quantize

java.lang.Object

de.sciss.jump3r.mp3.Quantize

public class Quantize
extends java.lang.Object

Constructor Summary

Constructors

Constructor and Description
Quantize()

Method Summary

Modifier and Type	Method and Description
void	bitpressure_strategy(LameInternalFlags gfc, float[][][] l3_xmin, int[][] min_bits, int[][] max_bits)
void	calc_target_bits(LameGlobalFlags gfp, float[][] pe, float[] ms_ener_ratio, int[][] targ_bits, int[] analog_silence_bits, int[] max_frame_bits) calculates target bits for ABR encoding mt 2000/05/31
void	get_framebits(LameGlobalFlags gfp, int[] frameBits) Robert Hegemann 2000-09-05 calculates * how many bits are available for analog silent granules * how many bits to use for the lowest allowed bitrate * how many bits each bitrate would provide
void	init_outer_loop(LameInternalFlags gfc, GrInfo cod_info)
boolean	init_xrpow(LameInternalFlags gfc, GrInfo cod_info, float[] xrpow)
void	iteration_finish_one(LameInternalFlags gfc, int gr, int ch) Robert Hegemann 2000-09-06 update reservoir status after FINAL quantization/bitrate
void	ms_convert(IIISideInfo l3_side, int gr) convert from L/R <.
int	outer_loop(LameGlobalFlags gfp, GrInfo cod_info, float[] l3_xmin, float[] xrpow, int ch, int targ_bits) Function: The outer iteration loop controls the masking conditions of all scalefactorbands.
void	setModules(BitStream bs, Reservoir rv, QuantizePVT qupvt, Takehiro tk)
void	trancate_smallspectrums(LameInternalFlags gfc, GrInfo gi, float[] l3_xmin, float[] work)
void	VBR_encode_granule(LameGlobalFlags gfp, GrInfo cod_info, float[] l3_xmin, float[] xrpow, int ch, int min_bits, int max_bits) 2000-09-04 Robert Hegemann
int	VBR_new_prepare(LameGlobalFlags gfp, float[][] pe, III_psy_ratio[][] ratio, float[][][] l3_xmin, int[] frameBits, int[][] max_bits)
int	VBR_old_prepare(LameGlobalFlags gfp, float[][] pe, float[] ms_ener_ratio, III_psy_ratio[][] ratio, float[][][] l3_xmin, int[] frameBits, int[][] min_bits, int[][] max_bits, int[][] bands) 2000-09-04 Robert Hegemann * converts LR to MS coding when necessary * calculates allowed/adjusted quantization noise amounts * detects analog silent frames some remarks: - lower masking depending on Quality setting - quality control together with adjusted ATH MDCT scaling on lower quality setting allocate more noise from ATH masking, and on higher quality setting allocate less noise from ATH masking.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

Quantize

public Quantize()

Method Detail

setModules

public final void setModules(BitStream bs,
                             Reservoir rv,
                             QuantizePVT qupvt,
                             Takehiro tk)

ms_convert

public final void ms_convert(IIISideInfo l3_side,
                             int gr)

convert from L/R <. Mid/Side

init_xrpow

public final boolean init_xrpow(LameInternalFlags gfc,
                                GrInfo cod_info,
                                float[] xrpow)

init_outer_loop

public final void init_outer_loop(LameInternalFlags gfc,
                                  GrInfo cod_info)

trancate_smallspectrums

public final void trancate_smallspectrums(LameInternalFlags gfc,
                                          GrInfo gi,
                                          float[] l3_xmin,
                                          float[] work)

outer_loop

public final int outer_loop(LameGlobalFlags gfp,
                            GrInfo cod_info,
                            float[] l3_xmin,
                            float[] xrpow,
                            int ch,
                            int targ_bits)

  Function: The outer iteration loop controls the masking conditions
  of all scalefactorbands. It computes the best scalefac and
  global gain. This module calls the inner iteration loop
 
  mt 5/99 completely rewritten to allow for bit reservoir control,
  mid/side channels with L/R or mid/side masking thresholds,
  and chooses best quantization instead of last quantization when
  no distortion free quantization can be found.
 
  added VBR support mt 5/99
 
  some code shuffle rh 9/00
 

Parameters:

l3_xmin - allowed distortion

xrpow - coloured magnitudes of spectral

targ_bits - maximum allowed bits

iteration_finish_one

public final void iteration_finish_one(LameInternalFlags gfc,
                                       int gr,
                                       int ch)

Robert Hegemann 2000-09-06 update reservoir status after FINAL quantization/bitrate

VBR_encode_granule

public final void VBR_encode_granule(LameGlobalFlags gfp,
                                     GrInfo cod_info,
                                     float[] l3_xmin,
                                     float[] xrpow,
                                     int ch,
                                     int min_bits,
                                     int max_bits)

2000-09-04 Robert Hegemann

Parameters:

l3_xmin - allowed distortion of the scalefactor

xrpow - coloured magnitudes of spectral values

get_framebits

public final void get_framebits(LameGlobalFlags gfp,
                                int[] frameBits)

Robert Hegemann 2000-09-05 calculates * how many bits are available for analog silent granules * how many bits to use for the lowest allowed bitrate * how many bits each bitrate would provide

VBR_old_prepare

public final int VBR_old_prepare(LameGlobalFlags gfp,
                                 float[][] pe,
                                 float[] ms_ener_ratio,
                                 III_psy_ratio[][] ratio,
                                 float[][][] l3_xmin,
                                 int[] frameBits,
                                 int[][] min_bits,
                                 int[][] max_bits,
                                 int[][] bands)

  2000-09-04 Robert Hegemann
 
  * converts LR to MS coding when necessary
  * calculates allowed/adjusted quantization noise amounts
  * detects analog silent frames
 
  some remarks:
  - lower masking depending on Quality setting
  - quality control together with adjusted ATH MDCT scaling
    on lower quality setting allocate more noise from
    ATH masking, and on higher quality setting allocate
    less noise from ATH masking.
  - experiments show that going more than 2dB over GPSYCHO's
    limits ends up in very annoying artefacts
 

bitpressure_strategy

public final void bitpressure_strategy(LameInternalFlags gfc,
                                       float[][][] l3_xmin,
                                       int[][] min_bits,
                                       int[][] max_bits)

VBR_new_prepare

public final int VBR_new_prepare(LameGlobalFlags gfp,
                                 float[][] pe,
                                 III_psy_ratio[][] ratio,
                                 float[][][] l3_xmin,
                                 int[] frameBits,
                                 int[][] max_bits)

calc_target_bits

public final void calc_target_bits(LameGlobalFlags gfp,
                                   float[][] pe,
                                   float[] ms_ener_ratio,
                                   int[][] targ_bits,
                                   int[] analog_silence_bits,
                                   int[] max_frame_bits)

calculates target bits for ABR encoding mt 2000/05/31