Added methods to Histogram and Distribution to get back the number of inserted samples and their total sum

This commit is contained in:
Paolo Cignoni 2014-05-13 10:54:42 +00:00
parent 5c7949d261
commit db0a706384
1 changed files with 244 additions and 304 deletions

View File

@ -8,7 +8,7 @@
* \ * * \ *
* All rights reserved. * * All rights reserved. *
* * * *
* This program is free software; you can redistribute it and/or modify * * This program is free software; you can redistribute it and/or modify *
* it under the terms of the GNU General Public License as published by * * it under the terms of the GNU General Public License as published by *
* the Free Software Foundation; either version 2 of the License, or * * the Free Software Foundation; either version 2 of the License, or *
* (at your option) any later version. * * (at your option) any later version. *
@ -20,70 +20,6 @@
* for more details. * * for more details. *
* * * *
****************************************************************************/ ****************************************************************************/
/****************************************************************************
History
$Log: not supported by cvs2svn $
Revision 1.18 2008/03/05 11:21:49 cignoni
Heavily refactored the whole structure.
Some interfaces have been changed. Be careful.
Revision 1.17 2008/02/29 12:15:06 cignoni
added maxcount
Revision 1.16 2006/11/28 21:29:21 cignoni
Re added typedef Histogramf and Histogramd
Revision 1.15 2006/11/28 09:47:42 corsini
add documentation
fix typo
Revision 1.14 2006/05/04 00:09:53 cignoni
minor change: removed unused vars
Revision 1.13 2006/03/29 09:25:47 zifnab1974
extra includes necessary for compilation of meshlab on AMD 64 with gcc 3.4.5
Revision 1.12 2006/01/12 13:12:54 callieri
in FileWrite, added file closing after finishing
Revision 1.11 2005/09/16 11:51:23 cignoni
removed signed/unsigned warning
Revision 1.10 2005/06/17 00:54:55 cignoni
Corrected small bug in SetRange (H was resized to n instead of n+1)
Revision 1.9 2005/06/14 14:27:00 ganovelli
added include of algorithm
Revision 1.8 2005/06/10 14:59:39 cignoni
Added include assert.h and cast to ScalarType for a pow in SetRange() function.
Revision 1.7 2005/06/09 14:19:55 cignoni
Added typedef Histogramf and Histogramd
Revision 1.6 2005/06/07 09:37:33 ponchio
Added fabs() to variance, which can sometime be negative in case
of rounding errors (and sqrt chokes on it).
Revision 1.5 2005/06/07 07:44:08 cignoni
Added Percentile and removed small bug in Add
Revision 1.4 2005/04/04 10:48:35 cignoni
Added missing functions Avg, rms etc, now fully (almost) functional
Revision 1.3 2005/03/14 09:23:40 cignoni
Added missing include<vector>
Revision 1.2 2004/08/25 15:15:26 ganovelli
minor changes to comply gcc compiler (typename's and stuff)
Revision 1.1 2004/06/24 09:12:28 cignoni
Initial Release
****************************************************************************/
#ifndef __VCG_HISTOGRAM #ifndef __VCG_HISTOGRAM
#define __VCG_HISTOGRAM #define __VCG_HISTOGRAM
@ -97,83 +33,86 @@ Initial Release
namespace vcg { namespace vcg {
template <class ScalarType> template <class ScalarType>
class Distribution class Distribution
{ {
private: private:
std::vector<ScalarType> vec; std::vector<ScalarType> vec;
bool dirty; bool dirty;
double valSum; double valSum;
double sqrdValSum; double sqrdValSum;
double avg; double avg;
double sqrdAvg; double sqrdAvg;
double rms; double rms;
double min_v; double min_v;
double max_v; double max_v;
public: public:
Distribution() { Clear(); } Distribution() { Clear(); }
void Clear() void Clear()
{ {
vec.clear(); vec.clear();
dirty=true; dirty=true;
min_v = std::numeric_limits<float>::max(); min_v = std::numeric_limits<float>::max();
max_v = -std::numeric_limits<float>::max(); max_v = -std::numeric_limits<float>::max();
} }
void Add(const ScalarType v)
{
vec.push_back(v);
dirty=true;
if(v<min_v) min_v=v;
if(v>max_v) max_v=v;
}
ScalarType Min() { return min_v; }
ScalarType Max() { return max_v; }
ScalarType Avg(){ DirtyCheck(); return avg;} void Add(const ScalarType v)
//! Returns the Root Mean Square of the data. {
ScalarType RMS(){ DirtyCheck(); return rms;} vec.push_back(v);
dirty=true;
//! Returns the variance of the data. if(v<min_v) min_v=v;
// the average of the squares less the square of the average. if(v>max_v) max_v=v;
ScalarType Variance(){ DirtyCheck(); return sqrdAvg - avg*avg ;} }
//! Returns the standard deviation of the data. ScalarType Min() const { return min_v; }
ScalarType StandardDeviation(){ DirtyCheck(); return sqrt( Variance() );} ScalarType Max() const { return max_v; }
void DirtyCheck() ScalarType Cnt() const { return ScalarType(vec.size()); }
{
if(!dirty) return; ScalarType Sum(){ DirtyCheck(); return valSum; }
std::sort(vec.begin(),vec.end()); ScalarType Avg(){ DirtyCheck(); return avg;}
valSum=0; //! Returns the Root Mean Square of the data.
sqrdValSum=0; ScalarType RMS(){ DirtyCheck(); return rms;}
typename std::vector<ScalarType>::iterator vi;
for(vi=vec.begin();vi!=vec.end();++vi) //! \brief Returns the variance of the data.
{ /// the average of the squares less the square of the average.
valSum += double(*vi); ScalarType Variance(){ DirtyCheck(); return sqrdAvg - avg*avg ;}
sqrdValSum += double(*vi)*double(*vi);
} //! Returns the standard deviation of the data.
avg = valSum/double(vec.size()); ScalarType StandardDeviation(){ DirtyCheck(); return sqrt( Variance() );}
sqrdAvg = sqrdValSum/double(vec.size());
rms = math::Sqrt(sqrdAvg); void DirtyCheck()
dirty=false; {
} if(!dirty) return;
std::sort(vec.begin(),vec.end());
ScalarType Percentile(ScalarType perc) valSum=0;
{ sqrdValSum=0;
assert(!vec.empty()); typename std::vector<ScalarType>::iterator vi;
assert(perc>=0 && perc<=1); for(vi=vec.begin();vi!=vec.end();++vi)
DirtyCheck(); {
int index = vec.size() *perc -1; valSum += double(*vi);
sqrdValSum += double(*vi)*double(*vi);
if(index< 0 ) index = 0; }
avg = valSum/double(vec.size());
return vec[index]; sqrdAvg = sqrdValSum/double(vec.size());
} rms = math::Sqrt(sqrdAvg);
dirty=false;
}
ScalarType Percentile(ScalarType perc)
{
assert(!vec.empty());
assert(perc>=0 && perc<=1);
DirtyCheck();
int index = vec.size() *perc -1;
if(index< 0 ) index = 0;
return vec[index];
}
}; };
@ -183,17 +122,17 @@ public:
* *
* This class implements a single-value histogram. * This class implements a single-value histogram.
*/ */
template <class ScalarType> template <class ScalarType>
class Histogram class Histogram
{ {
// public data members // public data members
protected: protected:
std::vector <ScalarType> H; //! Counters for bins. std::vector <ScalarType> H; //! Counters for bins.
std::vector <ScalarType> R; //! Range for bins. std::vector <ScalarType> R; //! Range for bins.
ScalarType minv; //! Minimum value. ScalarType minv; //! Minimum value.
ScalarType maxv; //! Maximum value. ScalarType maxv; //! Maximum value.
ScalarType minElem; //! Minimum value. ScalarType minElem; //! Minimum value.
ScalarType maxElem; //! Maximum value. ScalarType maxElem; //! Maximum value.
int n; //! Number of vaild intervals stored between minv and maxv. int n; //! Number of vaild intervals stored between minv and maxv.
@ -201,91 +140,92 @@ protected:
/// incrementally updated values /// incrementally updated values
ScalarType cnt; //! Number of accumulated samples. ScalarType cnt; //! Number of accumulated samples.
ScalarType avg; //! Average. ScalarType sum; //! Average.
ScalarType rms; //! Root mean square. ScalarType rms; //! Root mean square.
/** /**
* Returns the index of the bin which contains a given value. * Returns the index of the bin which contains a given value.
*/ */
int BinIndex(ScalarType val) ; int BinIndex(ScalarType val) ;
// public methods // public methods
public: public:
/**
* Set the histogram values.
*
* This method is used to correctly initialize the bins of the histogram.
* n is the number of valid intervals between minv and maxv.
* for a more robust working, the Histogram class stores also the two out of range intervals (-inf, minv] and [maxv, +inf)
* Each bin is left closed (eg it contains the value
* The \a gamma parameter is applied to modify the distribution of the ranges of the bins. Default uniform distibution.
*
*/
void SetRange(ScalarType _minv, ScalarType _maxv, int _n,ScalarType gamma=1.0 );
ScalarType MinV() {return minv;}; //! Minimum value. /**
ScalarType MaxV() {return maxv;}; //! Maximum value. * Set the histogram values.
*
* This method is used to correctly initialize the bins of the histogram.
* n is the number of valid intervals between minv and maxv.
* for a more robust working, the Histogram class stores also the two out of range intervals (-inf, minv] and [maxv, +inf)
* Each bin is left closed (eg it contains the value
* The \a gamma parameter is applied to modify the distribution of the ranges of the bins. Default uniform distibution.
*
*/
void SetRange(ScalarType _minv, ScalarType _maxv, int _n,ScalarType gamma=1.0 );
ScalarType MinV() {return minv;} //! Minimum value.
ScalarType MinElem() {return minElem;}; //! Minimum element added to the histogram. It could be < or > than MinV;. ScalarType MaxV() {return maxv;} //! Maximum value.
ScalarType MaxElem() {return maxElem;}; //! Maximum element added to the histogram. It could be < or > than MinV;.. ScalarType Sum() {return sum;} //! Total sum of inserted values.
ScalarType Cnt() {return cnt;}
/** ScalarType MinElem() {return minElem;} //! Minimum element added to the histogram. It could be < or > than MinV;.
* Add a new value to the histogram. ScalarType MaxElem() {return maxElem;} //! Maximum element added to the histogram. It could be < or > than MinV;..
*
* The statistics related to the histogram data (average, RMS, etc.) are /**
* also updated. * Add a new value to the histogram.
*/ *
* The statistics related to the histogram data (average, RMS, etc.) are
* also updated.
*/
void Add(ScalarType v, ScalarType increment=ScalarType(1.0)); void Add(ScalarType v, ScalarType increment=ScalarType(1.0));
ScalarType MaxCount() const; ScalarType MaxCount() const;
int BinNum() const {return n;}; int BinNum() const {return n;}
ScalarType BinCount(ScalarType v); ScalarType BinCount(ScalarType v);
ScalarType BinCountInd(int index) {return H[index];} ScalarType BinCountInd(int index) {return H[index];}
ScalarType BinCount(ScalarType v, ScalarType width); ScalarType BinCount(ScalarType v, ScalarType width);
ScalarType BinLowerBound(int index) {return R[index];} ScalarType BinLowerBound(int index) {return R[index];}
ScalarType BinUpperBound(int index) {return R[index+1];}; ScalarType BinUpperBound(int index) {return R[index+1];}
ScalarType RangeCount(ScalarType rangeMin, ScalarType rangeMax); ScalarType RangeCount(ScalarType rangeMin, ScalarType rangeMax);
ScalarType BinWidth(ScalarType v); ScalarType BinWidth(ScalarType v);
/** /**
* Returns the value corresponding to a given percentile of the data. * Returns the value corresponding to a given percentile of the data.
* *
* The percentile range between 0 and 1. * The percentile range between 0 and 1.
*/ */
ScalarType Percentile(ScalarType frac) const; ScalarType Percentile(ScalarType frac) const;
//! Returns the average of the data.
ScalarType Avg(){ return avg/cnt;}
//! Returns the Root Mean Square of the data.
ScalarType RMS(){ return sqrt(rms/double(cnt));}
//! Returns the variance of the data.
ScalarType Variance(){ return fabs(rms/cnt-Avg()*Avg());}
//! Returns the standard deviation of the data.
ScalarType StandardDeviation(){ return sqrt(Variance());}
//! Dump the histogram to a file. //! Returns the average of the data.
void FileWrite(const std::string &filename); ScalarType Avg(){ return sum/cnt;}
//! Reset histogram data. //! Returns the Root Mean Square of the data.
void Clear(); ScalarType RMS(){ return sqrt(rms/double(cnt));}
//! Returns the variance of the data.
ScalarType Variance(){ return fabs(rms/cnt-Avg()*Avg());}
//! Returns the standard deviation of the data.
ScalarType StandardDeviation(){ return sqrt(Variance());}
//! Dump the histogram to a file.
void FileWrite(const std::string &filename);
//! Reset histogram data.
void Clear();
}; };
template <class ScalarType> template <class ScalarType>
void Histogram<ScalarType>::Clear() void Histogram<ScalarType>::Clear()
{ {
H.clear(); H.clear();
R.clear(); R.clear();
cnt=0; cnt=0;
avg=0; sum=0;
rms=0; rms=0;
n=0; n=0;
minv=0; minv=0;
maxv=1; maxv=1;
minElem = std::numeric_limits<ScalarType>::max(); minElem = std::numeric_limits<ScalarType>::max();
maxElem = -std::numeric_limits<ScalarType>::max(); maxElem = -std::numeric_limits<ScalarType>::max();
} }
@ -293,169 +233,169 @@ void Histogram<ScalarType>::Clear()
/* /*
Note that the histogram holds <n> valid bins plus two semi-infinite bins. Note that the histogram holds <n> valid bins plus two semi-infinite bins.
R[0] = -inf R[0] = -inf
R[1] = minv R[1] = minv
R[n+1] = maxv R[n+1] = maxv
R[n+2] = +inf R[n+2] = +inf
Eg. SetRange(0, 10, 5) asks for 5 intervals covering the 0..10 range Eg. SetRange(0, 10, 5) asks for 5 intervals covering the 0..10 range
H[0] H[1] H[2] H[3] H[4] H[5] H[6] H[0] H[1] H[2] H[3] H[4] H[5] H[6]
-inf 0 2 4 6 8 10 +inf -inf 0 2 4 6 8 10 +inf
R[0] R[1] R[2] R[3] R[4] R[5] R[6] R[7] R[0] R[1] R[2] R[3] R[4] R[5] R[6] R[7]
*/ */
template <class ScalarType> template <class ScalarType>
void Histogram<ScalarType>::SetRange(ScalarType _minv, ScalarType _maxv, int _n, ScalarType gamma) void Histogram<ScalarType>::SetRange(ScalarType _minv, ScalarType _maxv, int _n, ScalarType gamma)
{ {
// reset data // reset data
Clear(); Clear();
minv=_minv;maxv=_maxv;n=_n;
H.resize(n+2);
fill(H.begin(),H.end(),0);
R.resize(n+3);
R[0] = - std::numeric_limits< ScalarType >::max();
R[n+2] = std::numeric_limits< ScalarType >::max();
double delta=(maxv-minv); minv=_minv;maxv=_maxv;n=_n;
if(gamma==1) H.resize(n+2);
{ fill(H.begin(),H.end(),0);
for(int i=0; i<=n; ++i) R.resize(n+3);
R[i+1] = minv + delta*ScalarType(i)/n;
} R[0] = - std::numeric_limits< ScalarType >::max();
else R[n+2] = std::numeric_limits< ScalarType >::max();
{
for(int i=0; i<=n; ++i) double delta=(maxv-minv);
R[i+1] = minv + delta*pow(ScalarType(i)/n,gamma); if(gamma==1)
} {
for(int i=0; i<=n; ++i)
R[i+1] = minv + delta*ScalarType(i)/n;
}
else
{
for(int i=0; i<=n; ++i)
R[i+1] = minv + delta*pow(ScalarType(i)/n,gamma);
}
} }
template <class ScalarType> template <class ScalarType>
int Histogram<ScalarType>::BinIndex(ScalarType val) int Histogram<ScalarType>::BinIndex(ScalarType val)
{ {
// lower_bound returns the furthermost iterator i in [first, last) such that, for every iterator j in [first, i), *j < value. // lower_bound returns the furthermost iterator i in [first, last) such that, for every iterator j in [first, i), *j < value.
// E.g. An iterator pointing to the first element "not less than" val, or end() if every element is less than val. // E.g. An iterator pointing to the first element "not less than" val, or end() if every element is less than val.
typename std::vector<ScalarType>::iterator it = lower_bound(R.begin(),R.end(),val); typename std::vector<ScalarType>::iterator it = lower_bound(R.begin(),R.end(),val);
assert(it!=R.begin()); assert(it!=R.begin());
assert(it!=R.end()); assert(it!=R.end());
assert((*it)>=val); assert((*it)>=val);
int pos = it-R.begin(); int pos = it-R.begin();
assert(pos >=1); assert(pos >=1);
pos -= 1; pos -= 1;
assert (R[pos] < val); assert (R[pos] < val);
assert ( val <= R[pos+1] ); assert ( val <= R[pos+1] );
return pos; return pos;
} }
/* /*
H[0] H[1] H[2] H[3] H[4] H[5] H[6] H[0] H[1] H[2] H[3] H[4] H[5] H[6]
-inf 0 2 4 6 8 10 +inf -inf 0 2 4 6 8 10 +inf
R[0] R[1] R[2] R[3] R[4] R[5] R[6] R[7] R[0] R[1] R[2] R[3] R[4] R[5] R[6] R[7]
asking for 3.14 lower bound will return an iterator pointing to R[3]==4; and will increase H[2] asking for 3.14 lower bound will return an iterator pointing to R[3]==4; and will increase H[2]
asking for 4 lower bound will return an iterator pointing to R[3]==4; and will increase H[2] asking for 4 lower bound will return an iterator pointing to R[3]==4; and will increase H[2]
*/ */
template <class ScalarType> template <class ScalarType>
void Histogram<ScalarType>::Add(ScalarType v, ScalarType increment) void Histogram<ScalarType>::Add(ScalarType v, ScalarType increment)
{ {
int pos=BinIndex(v); int pos=BinIndex(v);
if(v<minElem) minElem=v; if(v<minElem) minElem=v;
if(v>maxElem) maxElem=v; if(v>maxElem) maxElem=v;
assert((pos>=0)&&(pos<=n+1)); assert((pos>=0)&&(pos<=n+1));
H[pos]+=increment; H[pos]+=increment;
cnt+=increment; cnt+=increment;
avg+=v*increment; sum+=v*increment;
rms += (v*v)*increment; rms += (v*v)*increment;
} }
template <class ScalarType> template <class ScalarType>
ScalarType Histogram<ScalarType>::BinCount(ScalarType v) ScalarType Histogram<ScalarType>::BinCount(ScalarType v)
{ {
return H[BinIndex(v)]; return H[BinIndex(v)];
} }
template <class ScalarType> template <class ScalarType>
ScalarType Histogram<ScalarType>::BinCount(ScalarType v, ScalarType width) ScalarType Histogram<ScalarType>::BinCount(ScalarType v, ScalarType width)
{ {
return RangeCount(v-width/2.0,v+width/2.0); return RangeCount(v-width/2.0,v+width/2.0);
} }
template <class ScalarType> template <class ScalarType>
ScalarType Histogram<ScalarType>::RangeCount(ScalarType rangeMin, ScalarType rangeMax) ScalarType Histogram<ScalarType>::RangeCount(ScalarType rangeMin, ScalarType rangeMax)
{ {
int firstBin=BinIndex(rangeMin); int firstBin=BinIndex(rangeMin);
int lastBin=BinIndex (rangeMax); int lastBin=BinIndex (rangeMax);
ScalarType sum=0; ScalarType sum=0;
for(int i=firstBin; i<=lastBin;++i) for(int i=firstBin; i<=lastBin;++i)
sum+=H[i]; sum+=H[i];
return sum; return sum;
} }
template <class ScalarType> template <class ScalarType>
ScalarType Histogram<ScalarType>::BinWidth(ScalarType v) ScalarType Histogram<ScalarType>::BinWidth(ScalarType v)
{ {
int pos=BinIndex(v); int pos=BinIndex(v);
return R[pos+1]-R[pos]; return R[pos+1]-R[pos];
} }
template <class ScalarType> template <class ScalarType>
void Histogram<ScalarType>::FileWrite(const std::string &filename) void Histogram<ScalarType>::FileWrite(const std::string &filename)
{ {
FILE *fp; FILE *fp;
fp=fopen(filename.c_str(),"w"); fp=fopen(filename.c_str(),"w");
for(unsigned int i=0; i<H.size(); i++)
fprintf (fp,"%12.8lf , %12.8lf \n",R[i],double(H[i])/cnt);
fclose(fp); for(unsigned int i=0; i<H.size(); i++)
fprintf (fp,"%12.8lf , %12.8lf \n",R[i],double(H[i])/cnt);
fclose(fp);
} }
template <class ScalarType> template <class ScalarType>
ScalarType Histogram<ScalarType>::MaxCount() const ScalarType Histogram<ScalarType>::MaxCount() const
{ {
return *(std::max_element(H.begin(),H.end())); return *(std::max_element(H.begin(),H.end()));
} }
// Return the scalar value <r> such that there are <frac> samples <= <r>. // Return the scalar value <r> such that there are <frac> samples <= <r>.
// E.g. Percentile(0.0) will return R[1] e.g. min value // E.g. Percentile(0.0) will return R[1] e.g. min value
// E.g. Percentile(1.0) will return R[n+1] e.g max value // E.g. Percentile(1.0) will return R[n+1] e.g max value
template <class ScalarType> template <class ScalarType>
ScalarType Histogram<ScalarType>::Percentile(ScalarType frac) const ScalarType Histogram<ScalarType>::Percentile(ScalarType frac) const
{ {
if(H.size()==0 && R.size()==0) if(H.size()==0 && R.size()==0)
return 0; return 0;
// check percentile range
assert(frac >= 0 && frac <= 1);
ScalarType sum=0,partsum=0;
size_t i;
// useless summation just to be sure // check percentile range
for(i=0;i<H.size();i++) sum+=H[i]; assert(frac >= 0 && frac <= 1);
assert(sum==cnt);
sum*=frac; ScalarType sum=0,partsum=0;
for(i=0; i<H.size(); i++) size_t i;
{
partsum+=H[i]; // useless summation just to be sure
if(partsum>=sum) break; for(i=0;i<H.size();i++) sum+=H[i];
} assert(sum==cnt);
assert(i<H.size()); sum*=frac;
for(i=0; i<H.size(); i++)
return R[i+1]; {
partsum+=H[i];
if(partsum>=sum) break;
}
assert(i<H.size());
return R[i+1];
} }
typedef Histogram<double> Histogramd ; typedef Histogram<double> Histogramd ;