Dirac - A Video Codec

Created by the British Broadcasting Corporation.


wavelet_utils.h

Go to the documentation of this file.
00001 /* ***** BEGIN LICENSE BLOCK *****
00002 *
00003 * $Id: wavelet_utils.h,v 1.21 2006/10/05 15:22:52 tjdwave Exp $ $Name: Dirac_0_7_0 $
00004 *
00005 * Version: MPL 1.1/GPL 2.0/LGPL 2.1
00006 *
00007 * The contents of this file are subject to the Mozilla Public License
00008 * Version 1.1 (the "License"); you may not use this file except in compliance
00009 * with the License. You may obtain a copy of the License at
00010 * http://www.mozilla.org/MPL/
00011 *
00012 * Software distributed under the License is distributed on an "AS IS" basis,
00013 * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License for
00014 * the specific language governing rights and limitations under the License.
00015 *
00016 * The Original Code is BBC Research and Development code.
00017 *
00018 * The Initial Developer of the Original Code is the British Broadcasting
00019 * Corporation.
00020 * Portions created by the Initial Developer are Copyright (C) 2004.
00021 * All Rights Reserved.
00022 *
00023 * Contributor(s): Thomas Davies (Original Author),
00024 *                 Scott R Ladd
00025 *                 Anuradha Suraparaju
00026 *
00027 * Alternatively, the contents of this file may be used under the terms of
00028 * the GNU General Public License Version 2 (the "GPL"), or the GNU Lesser
00029 * Public License Version 2.1 (the "LGPL"), in which case the provisions of
00030 * the GPL or the LGPL are applicable instead of those above. If you wish to
00031 * allow use of your version of this file only under the terms of the either
00032 * the GPL or LGPL and not to allow others to use your version of this file
00033 * under the MPL, indicate your decision by deleting the provisions above
00034 * and replace them with the notice and other provisions required by the GPL
00035 * or LGPL. If you do not delete the provisions above, a recipient may use
00036 * your version of this file under the terms of any one of the MPL, the GPL
00037 * or the LGPL.
00038 * ***** END LICENSE BLOCK ***** */
00039 
00040 #ifndef _WAVELET_UTILS_H_
00041 #define _WAVELET_UTILS_H_
00042 
00043 #include <libdirac_common/arrays.h>
00044 #include <libdirac_common/common.h>
00045 #include <vector>
00046 #include <cmath>
00047 #include <iostream>
00048 
00049 //utilities for subband and wavelet transforms
00050 //Includes fast transform using lifting
00051 
00052 namespace dirac
00053 {
00054 
00055     class PicArray;
00056     class Subband;
00057 
00059     class CodeBlock
00060     {
00061 
00062     friend class Subband;
00063 
00064     public:
00066         /*
00067             Default constructor - sets all dimensions to zero
00068         */
00069         CodeBlock();
00070 
00072         /*
00073             Initialise the code block
00074             \param    xstart  the x-coord of the first coefficient in the block
00075             \param    xend    one past the last coefficient, horizontally    
00076             \param    ystart  the y-coord of the first coefficient in the block
00077             \param    yend    one past the last coefficient, vertically    
00078         */
00079         CodeBlock( const int xstart , const int ystart , const int xend , const int yend);
00080 
00082         int Xstart() const { return m_xstart; }
00083 
00085         int Ystart() const { return m_ystart; }
00086     
00088         int Xend() const { return m_xend; }
00089 
00091         int Yend() const { return m_yend; }
00092 
00094         int Xl() const { return m_xl; }
00095 
00097         int Yl() const { return m_yl; }
00098 
00100         int QIndex() const{ return m_qindex; }
00101 
00103         float Wt() const { return m_wt; }
00104 
00106         bool Skipped() const { return m_skipped; }
00107 
00109         void SetQIndex( const int qindex ){ m_qindex = qindex; }
00110 
00112         void SetSkip( bool skip ){ m_skipped = skip; }
00113 
00114     private:
00115 
00117         /*
00118             Initialise the code block
00119             \param    xstart  the x-coord of the first coefficient in the block
00120             \param    xend    one past the last coefficient, horizontally    
00121             \param    ystart  the y-coord of the first coefficient in the block
00122             \param    yend    one past the last coefficient, vertically    
00123         */
00124         void Init( const int xstart , const int ystart , const int xend , const int yend );
00125 
00127         void SetWt( const float w ){ m_wt = w; }
00128 
00129 
00130     private:
00131 
00132         int m_xstart;
00133         int m_ystart;
00134         int m_xend;
00135         int m_yend;
00136         int m_xl;
00137         int m_yl;
00138 
00139         int m_qindex;
00140         float m_wt;
00141 
00142         bool m_skipped;
00143     };
00144 
00145 
00147     class Subband
00148     {
00149     public:
00150 
00152         Subband();
00153 
00155 
00163         Subband(int xpos, int ypos, int xlen, int ylen);
00164 
00166 
00175         Subband(int xpos, int ypos, int xlen, int ylen, int d);
00176 
00178         ~Subband();
00179 
00180         //Default (shallow) copy constructor and operator= used
00181 
00183         int Xl() const {return m_xl;}
00184     
00186         int Xp() const {return m_xp;}
00187     
00189         int Yl() const {return m_yl;}
00190     
00192         int Yp() const {return m_yp;}
00193     
00195         int Max() const {return m_max_bit;}
00196     
00198         double Wt() const {return m_wt;}
00199     
00201         int Depth() const {return m_depth;}
00202     
00204         int Scale() const {return ( 1<<m_depth );}
00205     
00207         int QIndex() const {return m_qindex;}
00208 
00210         bool UsingMultiQuants() const {return m_multi_quants; } 
00211     
00213         int Parent() const {return m_parent;}
00214     
00216         const std::vector<int>& Children() const {return m_children;}
00217 
00219         int Child(const int n) const {return m_children[n];}
00220 
00222         TwoDArray<CodeBlock>& GetCodeBlocks(){ return m_code_block_array; } 
00223 
00225         const TwoDArray<CodeBlock>& GetCodeBlocks() const { return m_code_block_array; } 
00226 
00228         bool Skipped() const { return m_skipped; }
00229     
00231         void SetWt( const float w );
00232     
00234         void SetParent( const int p ){ m_parent=p; }
00235     
00237         void SetDepth( const int d ){ m_depth=d;}
00238     
00240         void SetMax( const int m ){ m_max_bit=m; };
00241     
00243         void SetChildren( const std::vector<int>& clist ){ m_children = clist; }
00244     
00246         void AddChild( const int c ){ m_children.push_back(c); }
00247 
00249         void SetNumBlocks( const int ynum , const int xnum );
00250 
00252         void SetQIndex( const int idx){ m_qindex = idx; }
00253 
00255         void SetUsingMultiQuants( const bool multi){ m_multi_quants = multi; }
00256 
00258         void SetSkip(const bool skip ){ m_skipped = skip; }
00259 
00260     private:
00261         // subband bounds
00262         int m_xp , m_yp , m_xl , m_yl; 
00263 
00264         // perceptual weight for quantisation
00265         double m_wt;
00266 
00267         // depth in the transform
00268         int m_depth;
00269 
00270         // quantiser index
00271         int m_qindex;
00272 
00273         // position of parent in a subband list
00274         int m_parent;
00275 
00276         // positions of children in the subband list
00277         std::vector<int> m_children;
00278 
00279         // position of the MSB of the largest absolute value
00280         int m_max_bit;              
00281 
00282         // The code blocks
00283         TwoDArray<CodeBlock> m_code_block_array;
00284 
00285         // A flag indicating whether we're using one qf for each code block
00286         bool m_multi_quants;
00287 
00288         // Whether the subband is skipped or not
00289         bool m_skipped;
00290     };
00291 
00293     class SubbandList
00294     {
00295     public:
00297         SubbandList(){}
00298 
00300         ~SubbandList(){}
00301 
00302         //Default (shallow) copy constructor and operator= used
00304         void Init(const int depth,const int xlen,const int ylen);
00305     
00307         int Length() const {return bands.size();}
00308     
00310         Subband& operator()(const int n){return bands[n-1];}
00311     
00313         const Subband& operator()(const int n) const {return bands[n-1];}    
00314     
00316         void AddBand(const Subband& b){bands.push_back(b);}
00317         
00319         void Clear(){bands.clear();}
00320     
00321     private:
00322 
00324         float PerceptualWeight( const float xf , const float yf , const CompSort cs);
00325 
00326     private:    
00327         std::vector<Subband> bands;
00328     };
00329 
00331 
00335     class WaveletTransform
00336     {
00337     public:
00339         WaveletTransform(int d = 4, WltFilter f = DAUB9_7);
00340         
00342         virtual ~WaveletTransform();
00343 
00345 
00350         void Transform(const Direction d, PicArray& pic_data);
00351     
00353         SubbandList& BandList(){return m_band_list;}
00354     
00356         const SubbandList& BandList() const {return m_band_list;}
00357     
00359 
00369         void SetBandWeights (const float cpd, 
00370                              const FrameSort& fsort,
00371                              const ChromaFormat& cformat,
00372                              const CompSort csort);
00373 
00374  
00375     private:
00376         // Classes used within wavelet transform
00377 
00379         class VHFilter
00380         {
00381 
00382         public:
00383 
00384             VHFilter(){}
00385 
00386             virtual ~VHFilter(){}
00387 
00389             virtual void Split(const int xp, const int yp, const int xl, const int yl, PicArray&pic_data)=0; 
00390 
00392             virtual void Synth(const int xp, const int yp, const int xl, const int yl, PicArray& pic_data)=0;
00393 
00395             virtual double GetLowFactor() const=0;
00396 
00398             virtual double GetHighFactor() const =0;
00399             
00401             virtual int GetShift() const =0;
00402 
00403         protected:
00404             
00406             inline void Interleave( const int xp, const int yp, const int xl, const int yl, PicArray&pic_data );
00407     
00408 
00410             inline void DeInterleave( const int xp, const int yp, const int xl, const int yl, PicArray&pic_data );
00411         
00413             void ShiftRowLeft(ValueType *row, int length, int shift);
00414         
00416             void ShiftRowRight(ValueType *row, int length, int shift);
00417         };
00418 
00420         class VHFilterDAUB9_7 : public VHFilter
00421         {
00422 
00423         public:
00424 
00426             void Split(const int xp, const int yp, const int xl, const int yl, PicArray&pic_data); 
00427 
00429             void Synth(const int xp, const int yp, const int xl, const int yl, PicArray& pic_data);
00430 
00432             double GetLowFactor() const { return 1.149604398;}
00433 
00435             double GetHighFactor() const { return 0.869864452;}
00436 
00438             int GetShift() const {return 1;}           
00439             
00440 
00441         };
00442 
00444         class VHFilterLEGALL5_3 : public VHFilter
00445         {
00446 
00447         public:
00448 
00450             void Split(const int xp, const int yp, const int xl, const int yl, PicArray&pic_data); 
00451 
00453             void Synth(const int xp, const int yp, const int xl, const int yl, PicArray& pic_data);
00454 
00456             double GetLowFactor() const { return 1.179535649;}    
00457 
00459             double GetHighFactor() const { return 0.81649658;}
00460 
00462             int GetShift() const {return 1;}           
00463 
00464 
00465 #ifdef HAVE_MMX
00466             inline void HorizSynth (int xp, int xl, int ystart, int yend, PicArray &pic_data);
00467 #endif
00468 
00469         };
00470 
00472         class VHFilterDD9_3 : public VHFilter
00473         {
00474 
00475         public:
00476 
00478             void Split(const int xp, const int yp, const int xl, const int yl, PicArray&pic_data); 
00479 
00481             void Synth(const int xp, const int yp, const int xl, const int yl, PicArray& pic_data);
00482 
00484             double GetLowFactor() const { return 1.218660804;}
00485 
00487             double GetHighFactor() const { return 0.780720058;}
00488             
00490             int GetShift() const {return 1;}                     
00491 
00492         };
00493 
00494 
00496         class VHFilterDD13_5 : public VHFilter
00497         {
00498 
00499         public:
00500 
00502             void Split(const int xp, const int yp, const int xl, const int yl, PicArray&pic_data); 
00503 
00505             void Synth(const int xp, const int yp, const int xl, const int yl, PicArray& pic_data);
00506 
00508             double GetLowFactor() const { return 1.28087;}
00509 
00511             double GetHighFactor() const { return 0.809254;}
00512             
00514             int GetShift() const {return 1;}           
00515             
00516         };
00517         
00519         class VHFilterHAAR0 : public VHFilter
00520         {
00521 
00522         public:
00523 
00525             void Split(const int xp, const int yp, const int xl, const int yl, PicArray&pic_data); 
00526 
00528             void Synth(const int xp, const int yp, const int xl, const int yl, PicArray& pic_data);
00529 
00531             double GetLowFactor() const { return 1.414213562;}    
00532 
00534             double GetHighFactor() const { return 0.707106781;}
00535 
00537             int GetShift() const {return 0;}
00538 
00539 
00540         };
00541 
00543         class VHFilterHAAR1 : public VHFilter
00544         {
00545 
00546         public:
00547 
00549             void Split(const int xp, const int yp, const int xl, const int yl, PicArray&pic_data); 
00550 
00552             void Synth(const int xp, const int yp, const int xl, const int yl, PicArray& pic_data);
00553 
00555             double GetLowFactor() const { return 1.414213562;}    
00556 
00558             double GetHighFactor() const { return 0.707106781;}
00559 
00561             int GetShift() const {return 1;}
00562 
00563 
00564         };
00565 
00566 
00568         class VHFilterHAAR2 : public VHFilter
00569         {
00570 
00571         public:
00572 
00574             void Split(const int xp, const int yp, const int xl, const int yl, PicArray&pic_data); 
00575 
00577             void Synth(const int xp, const int yp, const int xl, const int yl, PicArray& pic_data);
00578 
00580             double GetLowFactor() const { return 1.414213562;}    
00581 
00583             double GetHighFactor() const { return 0.707106781;}
00584             
00586             int GetShift() const {return 2;}            
00587 
00588         };
00589 
00590 
00591 
00592         // Lifting steps used in the filters
00593 
00595         template<int shift>
00596         class PredictStepShift
00597         {
00598 
00599         public:
00600 
00602             PredictStepShift(){}
00603 
00604             // Assume default copy constructor, assignment= and destructor //
00605 
00607             /*
00608                 Do the filtering.
00609                 \param   in_val   the value being predicted
00610                 \param   val1   the first value being used for prediction
00611                 \param   val2   the second value being used for prediction
00612             */
00613             inline void Filter(ValueType& in_val, const ValueType& val1, const ValueType& val2) const
00614             {
00615                 in_val -= (( val1 + val2 + (1<<(shift-1)) ) >>shift );
00616             }
00617 
00618         };
00619 
00621         template<int shift>
00622         class UpdateStepShift
00623         {
00624 
00625         public:
00627             UpdateStepShift(){}
00628 
00630             /*
00631                 Do the filtering.
00632                 \param   in_val   the value being updated
00633                 \param   val1   the first value being used for updating
00634                 \param   val2   the second value being used for updating
00635             */
00636             inline void Filter(ValueType& in_val, const ValueType& val1, const ValueType& val2) const
00637             {
00638                 in_val += ( ( val1 + val2 + (1<<(shift-1)) ) >>shift );
00639             }
00640 
00641         };  
00642 
00644         template <int shift , int tap1, int tap2>
00645         class PredictStepFourTap
00646         {
00647         public:
00648 
00650             PredictStepFourTap(){}
00651 
00652             // Assume default copy constructor, assignment= and destructor //
00653 
00655             inline void Filter(ValueType& in_val, const ValueType& val1, const ValueType& val2 ,
00656                                                   const ValueType& val3, const ValueType& val4 ) const
00657             {
00658                 in_val -= ( tap1*( val1 + val2 ) + tap2*( val3 + val4 ) + (1<<(shift-1)))>>shift;
00659             }
00660         }; 
00661 
00663         template <int shift , int tap1 , int tap2>
00664         class UpdateStepFourTap
00665         {
00666  
00667         public:
00669             UpdateStepFourTap(){}
00670 
00672             inline void Filter(ValueType& in_val, const ValueType& val1, const ValueType& val2 ,
00673                                                   const ValueType& val3, const ValueType& val4 ) const
00674             {
00675                 in_val += ( tap1*( val1 + val2 ) + tap2*( val3 + val4 ) + (1<<(shift-1)) )>>shift;
00676             }
00677         };  
00678 
00680         template <int gain> class PredictStep97
00681         {
00682         public:
00683 
00685             PredictStep97(){}
00686 
00687             // Assume default copy constructor, assignment= and destructor //
00688 
00690             /*
00691                 Do the filtering.
00692                 \param   in_val   the value being predicted
00693                 \param   val1   the first value being used for prediction
00694                 \param   val2   the second value being used for prediction
00695             */
00696             inline void Filter(ValueType& in_val, const ValueType& val1, const ValueType& val2) const
00697             {
00698                 in_val -= static_cast< ValueType >( (gain * static_cast< int >( val1 + val2 )) >>12 );
00699             }
00700         }; 
00701 
00703         template <int gain> class UpdateStep97
00704         {
00705  
00706         public:
00708             UpdateStep97(){}
00709 
00711             /*
00712                 Do the filtering.
00713                 \param   in_val   the value being updated
00714                 \param   val1   the first value being used for updating
00715                 \param   val2   the second value being used for updating
00716             */
00717             inline void Filter(ValueType& in_val, const ValueType& val1, const ValueType& val2) const
00718             {
00719                 in_val += static_cast< ValueType >( (gain * static_cast< int >( val1 + val2 )) >>12 );
00720             }
00721         };  
00722 
00723     private:
00724 
00725         // Private variables    
00726 
00727         SubbandList m_band_list;
00728 
00730         int m_depth;
00731     
00733         WltFilter m_filt_sort; 
00734 
00736         VHFilter* m_vhfilter;
00737 
00738     private:
00739         // Private functions
00741         WaveletTransform(const WaveletTransform& cpy);
00742     
00744         WaveletTransform& operator=(const WaveletTransform& rhs);
00745     
00747         float PerceptualWeight(float xf,float yf,CompSort cs);
00748    };  
00749 
00750 }// end namespace dirac
00751 
00752 #endif

© 2004 British Broadcasting Corporation. Dirac code licensed under the Mozilla Public License (MPL) Version 1.1.
HTML documentation generated by Dimitri van Heesch's excellent Doxygen tool.