From 5a6114e29975f66240ab7acae6e35fca65064a2c Mon Sep 17 00:00:00 2001
From: Alexandre Benoit <benoit.alexandre.vision@gmail.com>
Date: Fri, 31 Aug 2012 14:14:57 +0200
Subject: [PATCH] updated code for more flexible parallelisation : TBB parallel
 for loops are replaced by opencv parallel_for_ wrapper

---
 modules/contrib/src/basicretinafilter.cpp | 28 ++++++------
 modules/contrib/src/basicretinafilter.hpp | 54 +++++++++++------------
 modules/contrib/src/magnoretinafilter.cpp |  4 +-
 modules/contrib/src/magnoretinafilter.hpp | 24 +++++-----
 modules/contrib/src/parvoretinafilter.cpp |  4 +-
 modules/contrib/src/parvoretinafilter.hpp | 22 ++++-----
 modules/contrib/src/retinacolor.cpp       | 14 +++---
 modules/contrib/src/retinacolor.hpp       | 26 +++++------
 modules/contrib/src/templatebuffer.hpp    | 19 ++++----
 9 files changed, 97 insertions(+), 98 deletions(-)

diff --git a/modules/contrib/src/basicretinafilter.cpp b/modules/contrib/src/basicretinafilter.cpp
index a4270aa74d..c9e6a92e4d 100644
--- a/modules/contrib/src/basicretinafilter.cpp
+++ b/modules/contrib/src/basicretinafilter.cpp
@@ -345,8 +345,8 @@ void BasicRetinaFilter::_localLuminanceAdaptation(const float *inputFrame, const
 		//float tempMeanValue=meanLuminance+_meanInputValue*_tau;
 		updateCompressionParameter(meanLuminance);
 	}
-#ifdef HAVE_TBB
-        tbb::parallel_for(tbb::blocked_range<size_t>(0,_filterOutput.getNBpixels()), Parallel_localAdaptation(localLuminance, inputFrame, outputFrame, _localLuminanceFactor, _localLuminanceAddon, _maxInputValue), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL
+        cv::parallel_for_(cv::Range(0,_filterOutput.getNBpixels()), Parallel_localAdaptation(localLuminance, inputFrame, outputFrame, _localLuminanceFactor, _localLuminanceAddon, _maxInputValue));
 #else
 	//std::cout<<meanLuminance<<std::endl;
 	const float *localLuminancePTR=localLuminance;
@@ -466,8 +466,8 @@ void BasicRetinaFilter::_horizontalCausalFilter(float *outputFrame, unsigned int
 //  horizontal causal filter which adds the input inside
 void BasicRetinaFilter::_horizontalCausalFilter_addInput(const float *inputFrame, float *outputFrame, unsigned int IDrowStart, unsigned int IDrowEnd)
 {
-#ifdef HAVE_TBB
-        tbb::parallel_for(tbb::blocked_range<size_t>(IDrowStart,IDrowEnd), Parallel_horizontalCausalFilter_addInput(inputFrame, outputFrame, IDrowStart, _filterOutput.getNBcolumns(), _a, _tau), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL
+        cv::parallel_for_(cv::Range(IDrowStart,IDrowEnd), Parallel_horizontalCausalFilter_addInput(inputFrame, outputFrame, IDrowStart, _filterOutput.getNBcolumns(), _a, _tau));
 #else
 	for (unsigned int IDrow=IDrowStart; IDrow<IDrowEnd; ++IDrow)
 	{
@@ -487,8 +487,8 @@ void BasicRetinaFilter::_horizontalCausalFilter_addInput(const float *inputFrame
 void BasicRetinaFilter::_horizontalAnticausalFilter(float *outputFrame, unsigned int IDrowStart, unsigned int IDrowEnd)
 {
 
-#ifdef HAVE_TBB
-        tbb::parallel_for(tbb::blocked_range<size_t>(IDrowStart,IDrowEnd), Parallel_horizontalAnticausalFilter(outputFrame, IDrowEnd, _filterOutput.getNBcolumns(), _a ), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL
+        cv::parallel_for_(cv::Range(IDrowStart,IDrowEnd), Parallel_horizontalAnticausalFilter(outputFrame, IDrowEnd, _filterOutput.getNBcolumns(), _a ));
 #else
 	for (unsigned int IDrow=IDrowStart; IDrow<IDrowEnd; ++IDrow)
 	{
@@ -523,8 +523,8 @@ void BasicRetinaFilter::_horizontalAnticausalFilter_multGain(float *outputFrame,
 //  vertical anticausal filter
 void BasicRetinaFilter::_verticalCausalFilter(float *outputFrame, unsigned int IDcolumnStart, unsigned int IDcolumnEnd)
 {
-#ifdef HAVE_TBB
-        tbb::parallel_for(tbb::blocked_range<size_t>(IDcolumnStart,IDcolumnEnd), Parallel_verticalCausalFilter(outputFrame, _filterOutput.getNBrows(), _filterOutput.getNBcolumns(), _a ), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL
+        cv::parallel_for_(cv::Range(IDcolumnStart,IDcolumnEnd), Parallel_verticalCausalFilter(outputFrame, _filterOutput.getNBrows(), _filterOutput.getNBcolumns(), _a ));
 #else
         for (unsigned int IDcolumn=IDcolumnStart; IDcolumn<IDcolumnEnd; ++IDcolumn)
 	{
@@ -566,8 +566,8 @@ void BasicRetinaFilter::_verticalAnticausalFilter(float *outputFrame, unsigned i
 //  vertical anticausal filter which multiplies the output by _gain
 void BasicRetinaFilter::_verticalAnticausalFilter_multGain(float *outputFrame, unsigned int IDcolumnStart, unsigned int IDcolumnEnd)
 {
-#ifdef HAVE_TBB
-        tbb::parallel_for(tbb::blocked_range<size_t>(IDcolumnStart,IDcolumnEnd), Parallel_verticalAnticausalFilter_multGain(outputFrame, _filterOutput.getNBrows(), _filterOutput.getNBcolumns(), _a, _gain ), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL
+        cv::parallel_for_(cv::Range(IDcolumnStart,IDcolumnEnd), Parallel_verticalAnticausalFilter_multGain(outputFrame, _filterOutput.getNBrows(), _filterOutput.getNBcolumns(), _a, _gain ));
 #else
         float* offset=outputFrame+_filterOutput.getNBpixels()-_filterOutput.getNBcolumns();
 	//#pragma omp parallel for
@@ -819,8 +819,8 @@ void BasicRetinaFilter::_horizontalCausalFilter_Irregular_addInput(const float *
 //  horizontal anticausal filter  (basic way, no add on)
 void BasicRetinaFilter::_horizontalAnticausalFilter_Irregular(float *outputFrame, unsigned int IDrowStart, unsigned int IDrowEnd, const float *spatialConstantBuffer)
 {
-#ifdef HAVE_TBB
-        tbb::parallel_for(tbb::blocked_range<size_t>(IDrowStart,IDrowEnd), Parallel_horizontalAnticausalFilter_Irregular(outputFrame, spatialConstantBuffer, IDrowEnd, _filterOutput.getNBcolumns()), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL
+        cv::parallel_for_(cv::Range(IDrowStart,IDrowEnd), Parallel_horizontalAnticausalFilter_Irregular(outputFrame, spatialConstantBuffer, IDrowEnd, _filterOutput.getNBcolumns()));
 #else
 	register float* outputPTR=outputFrame+IDrowEnd*(_filterOutput.getNBcolumns())-1;
 	register const float* spatialConstantPTR=spatialConstantBuffer+IDrowEnd*(_filterOutput.getNBcolumns())-1;
@@ -841,8 +841,8 @@ void BasicRetinaFilter::_horizontalAnticausalFilter_Irregular(float *outputFrame
 //  vertical anticausal filter
 void BasicRetinaFilter::_verticalCausalFilter_Irregular(float *outputFrame, unsigned int IDcolumnStart, unsigned int IDcolumnEnd, const float *spatialConstantBuffer)
 {
-#ifdef HAVE_TBB
-        tbb::parallel_for(tbb::blocked_range<size_t>(IDcolumnStart,IDcolumnEnd), Parallel_verticalCausalFilter_Irregular(outputFrame, spatialConstantBuffer, _filterOutput.getNBrows(), _filterOutput.getNBcolumns()), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL
+        cv::parallel_for_(cv::Range(IDcolumnStart,IDcolumnEnd), Parallel_verticalCausalFilter_Irregular(outputFrame, spatialConstantBuffer, _filterOutput.getNBrows(), _filterOutput.getNBcolumns()));
 #else
 	for (unsigned int IDcolumn=IDcolumnStart; IDcolumn<IDcolumnEnd; ++IDcolumn)
 	{
diff --git a/modules/contrib/src/basicretinafilter.hpp b/modules/contrib/src/basicretinafilter.hpp
index c0cf100a4c..0039c9a6f2 100644
--- a/modules/contrib/src/basicretinafilter.hpp
+++ b/modules/contrib/src/basicretinafilter.hpp
@@ -436,16 +436,16 @@ protected:
 	void _local_verticalCausalFilter(float *outputFrame, unsigned int IDcolumnStart, unsigned int IDcolumnEnd, const unsigned int *integrationAreas);
 	void _local_verticalAnticausalFilter_multGain(float *outputFrame, unsigned int IDcolumnStart, unsigned int IDcolumnEnd, const unsigned int *integrationAreas); // this functions affects _gain at the output
 
-#ifdef HAVE_TBB
+#ifdef MAKE_PARALLEL
 /******************************************************
-** IF TBB is useable, then, main loops are parallelized using these functors
+** IF some parallelizing thread methods are available, then, main loops are parallelized using these functors
 ** ==> main idea paralellise main filters loops, then, only the most used methods are parallelized... TODO : increase the number of parallelised methods as necessary
 ** ==> functors names = Parallel_$$$ where $$$= the name of the serial method that is parallelised
 ** ==> functors constructors can differ from the parameters used with their related serial functions
 */
 
 #define _DEBUG_TBB // define DEBUG_TBB in order to display additionnal data on stdout
-    class Parallel_horizontalAnticausalFilter
+    class Parallel_horizontalAnticausalFilter: public cv::ParallelLoopBody
     {
     private:
         float *outputFrame;
@@ -465,16 +465,16 @@ protected:
 #endif
         }
 
-        void operator()( const tbb::blocked_range<size_t>& r ) const {
+        virtual void operator()( const Range& r ) const {
 
 #ifdef DEBUG_TBB
             std::cout<<"Parallel_horizontalAnticausalFilter::operator() :"
 		<<"\n\t range size="<<r.size()
-		<<"\n\t first index="<<r.begin()
+		<<"\n\t first index="<<r.start
 		//<<"\n\t last index="<<filterParam
 		<<std::endl;
 #endif
-            for (size_t IDrow=r.begin(); IDrow!=r.end(); ++IDrow)
+            for (int IDrow=r.start; IDrow!=r.end; ++IDrow)
 	    {
                 register float* outputPTR=outputFrame+(IDrowEnd-IDrow)*(nbColumns)-1;
 		register float result=0;
@@ -487,7 +487,7 @@ protected:
     }
     };
 
-    class Parallel_horizontalCausalFilter_addInput
+    class Parallel_horizontalCausalFilter_addInput: public cv::ParallelLoopBody
     {
     private:
         const float *inputFrame;
@@ -498,8 +498,8 @@ protected:
 	Parallel_horizontalCausalFilter_addInput(const float *bufferToAddAsInputProcess, float *bufferToProcess, const unsigned int idStart, const unsigned int nbCols,  const float a,  const float tau)
         :inputFrame(bufferToAddAsInputProcess), outputFrame(bufferToProcess), IDrowStart(idStart), nbColumns(nbCols), filterParam_a(a), filterParam_tau(tau){}
 
-        void operator()( const tbb::blocked_range<size_t>& r ) const {
-	    for (unsigned int IDrow=r.begin(); IDrow!=r.end(); ++IDrow)
+        virtual void operator()( const Range& r ) const {
+	    for (int IDrow=r.start; IDrow!=r.end; ++IDrow)
 	    {
 		register float* outputPTR=outputFrame+(IDrowStart+IDrow)*nbColumns;
 		register const float* inputPTR=inputFrame+(IDrowStart+IDrow)*nbColumns;
@@ -513,7 +513,7 @@ protected:
         }
     };
 
-    class Parallel_verticalCausalFilter
+    class Parallel_verticalCausalFilter: public cv::ParallelLoopBody
     {
     private:
         float *outputFrame;
@@ -523,8 +523,8 @@ protected:
         Parallel_verticalCausalFilter(float *bufferToProcess, const unsigned int nbRws, const unsigned int nbCols, const float a )
         :outputFrame(bufferToProcess), nbRows(nbRws), nbColumns(nbCols), filterParam_a(a){}
         
-        void operator()( const tbb::blocked_range<size_t>& r ) const {
-            for (unsigned int IDcolumn=r.begin(); IDcolumn!=r.end(); ++IDcolumn)
+        virtual void operator()( const Range& r ) const {
+            for (int IDcolumn=r.start; IDcolumn!=r.end; ++IDcolumn)
 	    {
 		register float result=0;
 		register float *outputPTR=outputFrame+IDcolumn;
@@ -540,7 +540,7 @@ protected:
         }
     };
 
-    class Parallel_verticalAnticausalFilter_multGain
+    class Parallel_verticalAnticausalFilter_multGain: public cv::ParallelLoopBody
     {
     private:
         float *outputFrame;
@@ -550,9 +550,9 @@ protected:
         Parallel_verticalAnticausalFilter_multGain(float *bufferToProcess, const unsigned int nbRws, const unsigned int nbCols, const float a, const float  gain)
         :outputFrame(bufferToProcess), nbRows(nbRws), nbColumns(nbCols), filterParam_a(a), filterParam_gain(gain){}
         
-        void operator()( const tbb::blocked_range<size_t>& r ) const {
+        virtual void operator()( const Range& r ) const {
             float* offset=outputFrame+nbColumns*nbRows-nbColumns;
-    	    for (unsigned int IDcolumn=r.begin(); IDcolumn!=r.end(); ++IDcolumn)
+    	    for (int IDcolumn=r.start; IDcolumn!=r.end; ++IDcolumn)
 	    {
 		register float result=0;
 		register float *outputPTR=offset+IDcolumn;
@@ -568,7 +568,7 @@ protected:
         }
     };
 
-    class Parallel_localAdaptation
+    class Parallel_localAdaptation: public cv::ParallelLoopBody
     {
     private:
          const float *localLuminance, *inputFrame;
@@ -578,11 +578,11 @@ protected:
          Parallel_localAdaptation(const float *localLum, const float *inputImg, float *bufferToProcess, const float localLuminanceFact, const float localLuminanceAdd, const float maxInputVal)
          :localLuminance(localLum), inputFrame(inputImg),outputFrame(bufferToProcess), localLuminanceFactor(localLuminanceFact), localLuminanceAddon(localLuminanceAdd), maxInputValue(maxInputVal) {};
 
-         void operator()( const tbb::blocked_range<size_t>& r ) const {
-             const float *localLuminancePTR=localLuminance+r.begin();
-	     const float *inputFramePTR=inputFrame+r.begin();
-	     float *outputFramePTR=outputFrame+r.begin();
-	     for (register unsigned int IDpixel=r.begin() ; IDpixel!=r.end() ; ++IDpixel, ++inputFramePTR, ++outputFramePTR)
+         virtual void operator()( const Range& r ) const {
+             const float *localLuminancePTR=localLuminance+r.start;
+	     const float *inputFramePTR=inputFrame+r.start;
+	     float *outputFramePTR=outputFrame+r.start;
+	     for (register int IDpixel=r.start ; IDpixel!=r.end ; ++IDpixel, ++inputFramePTR, ++outputFramePTR)
 	     {
 		float X0=*(localLuminancePTR++)*localLuminanceFactor+localLuminanceAddon;
 		// TODO : the following line can lead to a divide by zero ! A small offset is added, take care if the offset is too large in case of High Dynamic Range images which can use very small values...		
@@ -594,7 +594,7 @@ protected:
 
 //////////////////////////////////////////
 /// Specific filtering methods which manage non const spatial filtering parameter (used By retinacolor and LogProjectors) 
-    class Parallel_horizontalAnticausalFilter_Irregular
+    class Parallel_horizontalAnticausalFilter_Irregular: public cv::ParallelLoopBody
     {
     private:
 	float *outputFrame;
@@ -604,9 +604,9 @@ protected:
         Parallel_horizontalAnticausalFilter_Irregular(float *bufferToProcess, const float *spatialConst, const unsigned int idEnd, const unsigned int nbCols)
         :outputFrame(bufferToProcess), spatialConstantBuffer(spatialConst), IDrowEnd(idEnd), nbColumns(nbCols){}
 
-void operator()( const tbb::blocked_range<size_t>& r ) const {
+virtual void operator()( const Range& r ) const {
 
-            for (size_t IDrow=r.begin(); IDrow!=r.end(); ++IDrow)
+            for (int IDrow=r.start; IDrow!=r.end; ++IDrow)
 	    {
                 register float* outputPTR=outputFrame+(IDrowEnd-IDrow)*(nbColumns)-1;
                 register const float* spatialConstantPTR=spatialConstantBuffer+(IDrowEnd-IDrow)*(nbColumns)-1;
@@ -620,7 +620,7 @@ void operator()( const tbb::blocked_range<size_t>& r ) const {
         }
     };
 
-    class Parallel_verticalCausalFilter_Irregular
+    class Parallel_verticalCausalFilter_Irregular: public cv::ParallelLoopBody
     {
     private:
         float *outputFrame;
@@ -630,8 +630,8 @@ void operator()( const tbb::blocked_range<size_t>& r ) const {
         Parallel_verticalCausalFilter_Irregular(float *bufferToProcess, const float *spatialConst, const unsigned int nbRws, const unsigned int nbCols)
         :outputFrame(bufferToProcess), spatialConstantBuffer(spatialConst), nbRows(nbRws), nbColumns(nbCols){}
         
-        void operator()( const tbb::blocked_range<size_t>& r ) const {
-            for (unsigned int IDcolumn=r.begin(); IDcolumn!=r.end(); ++IDcolumn)
+        virtual void operator()( const Range& r ) const {
+            for (int IDcolumn=r.start; IDcolumn!=r.end; ++IDcolumn)
 	    {
 		register float result=0;
 		register float *outputPTR=outputFrame+IDcolumn;
diff --git a/modules/contrib/src/magnoretinafilter.cpp b/modules/contrib/src/magnoretinafilter.cpp
index 4626133efa..9bdf4e613f 100644
--- a/modules/contrib/src/magnoretinafilter.cpp
+++ b/modules/contrib/src/magnoretinafilter.cpp
@@ -153,8 +153,8 @@ void MagnoRetinaFilter::setCoefficientsTable(const float parasolCells_beta, cons
 
 void MagnoRetinaFilter::_amacrineCellsComputing(const float *OPL_ON, const float *OPL_OFF)
 {
-#ifdef HAVE_TBB
-        tbb::parallel_for(tbb::blocked_range<size_t>(0,_filterOutput.getNBpixels()), Parallel_amacrineCellsComputing(OPL_ON, OPL_OFF, &_previousInput_ON[0], &_previousInput_OFF[0], &_amacrinCellsTempOutput_ON[0], &_amacrinCellsTempOutput_OFF[0], _temporalCoefficient), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL
+        cv::parallel_for_(cv::Range(0,_filterOutput.getNBpixels()), Parallel_amacrineCellsComputing(OPL_ON, OPL_OFF, &_previousInput_ON[0], &_previousInput_OFF[0], &_amacrinCellsTempOutput_ON[0], &_amacrinCellsTempOutput_OFF[0], _temporalCoefficient));
 #else
 	register const float *OPL_ON_PTR=OPL_ON;
 	register const float *OPL_OFF_PTR=OPL_OFF;
diff --git a/modules/contrib/src/magnoretinafilter.hpp b/modules/contrib/src/magnoretinafilter.hpp
index daefb74485..b14abfd5c6 100644
--- a/modules/contrib/src/magnoretinafilter.hpp
+++ b/modules/contrib/src/magnoretinafilter.hpp
@@ -192,14 +192,14 @@ private:
 
 	// amacrine cells filter : high pass temporal filter
 	void _amacrineCellsComputing(const float *ONinput, const float *OFFinput);
-#ifdef HAVE_TBB
+#ifdef MAKE_PARALLEL
 /******************************************************
-** IF TBB is useable, then, main loops are parallelized using these functors
+** IF some parallelizing thread methods are available, then, main loops are parallelized using these functors
 ** ==> main idea paralellise main filters loops, then, only the most used methods are parallelized... TODO : increase the number of parallelised methods as necessary
 ** ==> functors names = Parallel_$$$ where $$$= the name of the serial method that is parallelised
 ** ==> functors constructors can differ from the parameters used with their related serial functions
 */
-    class Parallel_amacrineCellsComputing
+    class Parallel_amacrineCellsComputing: public cv::ParallelLoopBody
     {
     private:
 	const float *OPL_ON, *OPL_OFF;
@@ -209,15 +209,15 @@ private:
         Parallel_amacrineCellsComputing(const float *OPL_ON_PTR, const float *OPL_OFF_PTR, float *previousInput_ON_PTR, float *previousInput_OFF_PTR, float *amacrinCellsTempOutput_ON_PTR, float *amacrinCellsTempOutput_OFF_PTR, float temporalCoefficientVal)
         :OPL_ON(OPL_ON_PTR), OPL_OFF(OPL_OFF_PTR), previousInput_ON(previousInput_ON_PTR), previousInput_OFF(previousInput_OFF_PTR), amacrinCellsTempOutput_ON(amacrinCellsTempOutput_ON_PTR), amacrinCellsTempOutput_OFF(amacrinCellsTempOutput_OFF_PTR), temporalCoefficient(temporalCoefficientVal) {}
         
-        void operator()( const tbb::blocked_range<size_t>& r ) const {
-	register const float *OPL_ON_PTR=OPL_ON+r.begin();
-	register const float *OPL_OFF_PTR=OPL_OFF+r.begin();
-	register float *previousInput_ON_PTR= previousInput_ON+r.begin();
-	register float *previousInput_OFF_PTR= previousInput_OFF+r.begin();
-	register float *amacrinCellsTempOutput_ON_PTR= amacrinCellsTempOutput_ON+r.begin();
-	register float *amacrinCellsTempOutput_OFF_PTR= amacrinCellsTempOutput_OFF+r.begin();
-
-	for (unsigned int IDpixel=r.begin() ; IDpixel!=r.end(); ++IDpixel)
+        virtual void operator()( const Range& r ) const {
+	register const float *OPL_ON_PTR=OPL_ON+r.start;
+	register const float *OPL_OFF_PTR=OPL_OFF+r.start;
+	register float *previousInput_ON_PTR= previousInput_ON+r.start;
+	register float *previousInput_OFF_PTR= previousInput_OFF+r.start;
+	register float *amacrinCellsTempOutput_ON_PTR= amacrinCellsTempOutput_ON+r.start;
+	register float *amacrinCellsTempOutput_OFF_PTR= amacrinCellsTempOutput_OFF+r.start;
+
+	for (int IDpixel=r.start ; IDpixel!=r.end; ++IDpixel)
 	{
 
 		/* Compute ON and OFF amacrin cells high pass temporal filter */
diff --git a/modules/contrib/src/parvoretinafilter.cpp b/modules/contrib/src/parvoretinafilter.cpp
index 50d1b0ba68..64464ebddb 100644
--- a/modules/contrib/src/parvoretinafilter.cpp
+++ b/modules/contrib/src/parvoretinafilter.cpp
@@ -204,8 +204,8 @@ void ParvoRetinaFilter::_OPL_OnOffWaysComputing() // WARNING : this method requi
 	// loop that makes the difference between photoreceptor cells output and horizontal cells
 	// positive part goes on the ON way, negative pat goes on the OFF way
 
-#ifdef HAVE_TBB
-        tbb::parallel_for(tbb::blocked_range<size_t>(0,_filterOutput.getNBpixels()), Parallel_OPL_OnOffWaysComputing(&_photoreceptorsOutput[0], &_horizontalCellsOutput[0], &_bipolarCellsOutputON[0], &_bipolarCellsOutputOFF[0], &_parvocellularOutputON[0], &_parvocellularOutputOFF[0]), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL
+        cv::parallel_for_(cv::Range(0,_filterOutput.getNBpixels()), Parallel_OPL_OnOffWaysComputing(&_photoreceptorsOutput[0], &_horizontalCellsOutput[0], &_bipolarCellsOutputON[0], &_bipolarCellsOutputOFF[0], &_parvocellularOutputON[0], &_parvocellularOutputOFF[0]));
 #else
 	float *photoreceptorsOutput_PTR= &_photoreceptorsOutput[0];
 	float *horizontalCellsOutput_PTR= &_horizontalCellsOutput[0];
diff --git a/modules/contrib/src/parvoretinafilter.hpp b/modules/contrib/src/parvoretinafilter.hpp
index 76f5506855..2131ef9162 100644
--- a/modules/contrib/src/parvoretinafilter.hpp
+++ b/modules/contrib/src/parvoretinafilter.hpp
@@ -216,14 +216,14 @@ private:
 	// private functions
 	void _OPL_OnOffWaysComputing();
 
-#ifdef HAVE_TBB
+#ifdef MAKE_PARALLEL
 /******************************************************
-** IF TBB is useable, then, main loops are parallelized using these functors
+** IF some parallelizing thread methods are available, then, main loops are parallelized using these functors
 ** ==> main idea paralellise main filters loops, then, only the most used methods are parallelized... TODO : increase the number of parallelised methods as necessary
 ** ==> functors names = Parallel_$$$ where $$$= the name of the serial method that is parallelised
 ** ==> functors constructors can differ from the parameters used with their related serial functions
 */
-    class Parallel_OPL_OnOffWaysComputing
+    class Parallel_OPL_OnOffWaysComputing: public cv::ParallelLoopBody
     {
     private:
 	float *photoreceptorsOutput, *horizontalCellsOutput, *bipolarCellsON, *bipolarCellsOFF, *parvocellularOutputON, *parvocellularOutputOFF;
@@ -231,17 +231,17 @@ private:
         Parallel_OPL_OnOffWaysComputing(float *photoreceptorsOutput_PTR, float *horizontalCellsOutput_PTR, float *bipolarCellsON_PTR, float *bipolarCellsOFF_PTR, float *parvocellularOutputON_PTR, float *parvocellularOutputOFF_PTR)
         :photoreceptorsOutput(photoreceptorsOutput_PTR), horizontalCellsOutput(horizontalCellsOutput_PTR), bipolarCellsON(bipolarCellsON_PTR), bipolarCellsOFF(bipolarCellsOFF_PTR), parvocellularOutputON(parvocellularOutputON_PTR), parvocellularOutputOFF(parvocellularOutputOFF_PTR) {}
         
-        void operator()( const tbb::blocked_range<size_t>& r ) const {
+        virtual void operator()( const Range& r ) const {
 	    // compute bipolar cells response equal to photoreceptors minus horizontal cells response
 	    // and copy the result on parvo cellular outputs... keeping time before their local contrast adaptation for final result
-	    float *photoreceptorsOutput_PTR= photoreceptorsOutput+r.begin();
-	    float *horizontalCellsOutput_PTR= horizontalCellsOutput+r.begin();
-	    float *bipolarCellsON_PTR = bipolarCellsON+r.begin();
-	    float *bipolarCellsOFF_PTR = bipolarCellsOFF+r.begin();
-	    float *parvocellularOutputON_PTR= parvocellularOutputON+r.begin();
-	    float *parvocellularOutputOFF_PTR= parvocellularOutputOFF+r.begin();
+	    float *photoreceptorsOutput_PTR= photoreceptorsOutput+r.start;
+	    float *horizontalCellsOutput_PTR= horizontalCellsOutput+r.start;
+	    float *bipolarCellsON_PTR = bipolarCellsON+r.start;
+	    float *bipolarCellsOFF_PTR = bipolarCellsOFF+r.start;
+	    float *parvocellularOutputON_PTR= parvocellularOutputON+r.start;
+	    float *parvocellularOutputOFF_PTR= parvocellularOutputOFF+r.start;
 
-            for (register unsigned int IDpixel=r.begin() ; IDpixel!=r.end() ; ++IDpixel)
+            for (register int IDpixel=r.start ; IDpixel!=r.end ; ++IDpixel)
 	    {
 		float pixelDifference = *(photoreceptorsOutput_PTR++) -*(horizontalCellsOutput_PTR++);
 		// test condition to allow write pixelDifference in ON or OFF buffer and 0 in the over
diff --git a/modules/contrib/src/retinacolor.cpp b/modules/contrib/src/retinacolor.cpp
index 247234ffef..d138582ce4 100644
--- a/modules/contrib/src/retinacolor.cpp
+++ b/modules/contrib/src/retinacolor.cpp
@@ -433,8 +433,8 @@ void RetinaColor::clipRGBOutput_0_maxInputValue(float *inputOutputBuffer, const
     if (inputOutputBuffer==NULL)
         inputOutputBuffer= &_demultiplexedColorFrame[0];
 
-#ifdef HAVE_TBB // call the TemplateBuffer TBB clipping method
-        tbb::parallel_for(tbb::blocked_range<size_t>(0,_filterOutput.getNBpixels()*3), Parallel_clipBufferValues<float>(inputOutputBuffer, 0, maxInputValue), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL // call the TemplateBuffer TBB clipping method
+        cv::parallel_for_(cv::Range(0,_filterOutput.getNBpixels()*3), Parallel_clipBufferValues<float>(inputOutputBuffer, 0,  maxInputValue));
 #else
     register float *inputOutputBufferPTR=inputOutputBuffer;
     for (register unsigned int jf = 0; jf < _filterOutput.getNBpixels()*3; ++jf, ++inputOutputBufferPTR)
@@ -580,8 +580,8 @@ void RetinaColor::_adaptiveSpatialLPfilter(const float *inputFrame, float *outpu
 //  horizontal causal filter which adds the input inside... replaces the parent _horizontalCausalFilter_Irregular_addInput by avoiding a product for each pixel
 void RetinaColor::_adaptiveHorizontalCausalFilter_addInput(const float *inputFrame, float *outputFrame, unsigned int IDrowStart, unsigned int IDrowEnd)
 {
-#ifdef HAVE_TBB
-        tbb::parallel_for(tbb::blocked_range<size_t>(IDrowStart,IDrowEnd), Parallel_adaptiveHorizontalCausalFilter_addInput(inputFrame, outputFrame, &_imageGradient[0], _filterOutput.getNBcolumns()), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL
+        cv::parallel_for_(cv::Range(IDrowStart,IDrowEnd), Parallel_adaptiveHorizontalCausalFilter_addInput(inputFrame, outputFrame, &_imageGradient[0], _filterOutput.getNBcolumns()));
 #else
     register float* outputPTR=outputFrame+IDrowStart*_filterOutput.getNBcolumns();
     register const float* inputPTR=inputFrame+IDrowStart*_filterOutput.getNBcolumns();
@@ -604,8 +604,8 @@ void RetinaColor::_adaptiveHorizontalCausalFilter_addInput(const float *inputFra
 //  vertical anticausal filter which multiplies the output by _gain... replaces the parent _verticalAnticausalFilter_multGain by avoiding a product for each pixel and taking into account the second layer of the _imageGradient buffer
 void RetinaColor::_adaptiveVerticalAnticausalFilter_multGain(float *outputFrame, unsigned int IDcolumnStart, unsigned int IDcolumnEnd)
 {
-#ifdef HAVE_TBB
-        tbb::parallel_for(tbb::blocked_range<size_t>(IDcolumnStart,IDcolumnEnd), Parallel_adaptiveVerticalAnticausalFilter_multGain(outputFrame, &_imageGradient[0]+_filterOutput.getNBpixels(), _filterOutput.getNBrows(), _filterOutput.getNBcolumns(), _gain), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL
+        cv::parallel_for_(cv::Range(IDcolumnStart,IDcolumnEnd), Parallel_adaptiveVerticalAnticausalFilter_multGain(outputFrame, &_imageGradient[0]+_filterOutput.getNBpixels(), _filterOutput.getNBrows(), _filterOutput.getNBcolumns(), _gain));
 #else
     float* outputOffset=outputFrame+_filterOutput.getNBpixels()-_filterOutput.getNBcolumns();
     float* gradOffset= &_imageGradient[0]+_filterOutput.getNBpixels()*2-_filterOutput.getNBcolumns();
@@ -661,8 +661,8 @@ void RetinaColor::_computeGradient(const float *luminance)
             }
         }
     }
-
 }
+
 bool RetinaColor::applyKrauskopfLMS2Acr1cr2Transform(std::valarray<float> &result)
 {
     bool processSuccess=true;
diff --git a/modules/contrib/src/retinacolor.hpp b/modules/contrib/src/retinacolor.hpp
index ec5d060c56..67834629e3 100644
--- a/modules/contrib/src/retinacolor.hpp
+++ b/modules/contrib/src/retinacolor.hpp
@@ -256,16 +256,16 @@ protected:
 	// color space transform
 	void _applyImageColorSpaceConversion(const std::valarray<float> &inputFrame, std::valarray<float> &outputFrame, const float *transformTable);
 
-#ifdef HAVE_TBB
+#ifdef MAKE_PARALLEL
 /******************************************************
-** IF TBB is useable, then, main loops are parallelized using these functors
+** IF some parallelizing thread methods are available, then, main loops are parallelized using these functors
 ** ==> main idea paralellise main filters loops, then, only the most used methods are parallelized... TODO : increase the number of parallelised methods as necessary
 ** ==> functors names = Parallel_$$$ where $$$= the name of the serial method that is parallelised
 ** ==> functors constructors can differ from the parameters used with their related serial functions
 */
 
 /* Template :
-    class 
+    class Parallel_ : public cv::ParallelLoopBody
     {
     private:
 
@@ -273,12 +273,12 @@ protected:
          Parallel_()
          : {}
     
-         void operator()( const tbb::blocked_range<size_t>& r ) const {
+         virtual void operator()( const cv::Range& r ) const {
 
         }
     }:
 */
-    class Parallel_adaptiveHorizontalCausalFilter_addInput
+    class Parallel_adaptiveHorizontalCausalFilter_addInput: public cv::ParallelLoopBody
     {
     private:
 	float *outputFrame;
@@ -288,11 +288,11 @@ protected:
          Parallel_adaptiveHorizontalCausalFilter_addInput(const float *inputImg, float *bufferToProcess, const float *imageGrad, const unsigned int nbCols)
          :outputFrame(bufferToProcess), inputFrame(inputImg), imageGradient(imageGrad), nbColumns(nbCols) {};
     
-         void operator()( const tbb::blocked_range<size_t>& r ) const {
-            register float* outputPTR=outputFrame+r.begin()*nbColumns;
-	    register const float* inputPTR=inputFrame+r.begin()*nbColumns;
-	    register const float *imageGradientPTR= imageGradient+r.begin()*nbColumns;
-	    for (unsigned int IDrow=r.begin(); IDrow!=r.end(); ++IDrow)
+         virtual void operator()( const Range& r ) const {
+            register float* outputPTR=outputFrame+r.start*nbColumns;
+	    register const float* inputPTR=inputFrame+r.start*nbColumns;
+	    register const float *imageGradientPTR= imageGradient+r.start*nbColumns;
+	    for (int IDrow=r.start; IDrow!=r.end; ++IDrow)
 	    {
 		register float result=0;
 		for (unsigned int index=0; index<nbColumns; ++index)
@@ -304,7 +304,7 @@ protected:
         }
     };
 
-    class Parallel_adaptiveVerticalAnticausalFilter_multGain
+    class Parallel_adaptiveVerticalAnticausalFilter_multGain: public cv::ParallelLoopBody
     {
     private:
         float *outputFrame;
@@ -315,10 +315,10 @@ protected:
         Parallel_adaptiveVerticalAnticausalFilter_multGain(float *bufferToProcess, const float *imageGrad, const unsigned int nbRws, const unsigned int nbCols, const float  gain)
         :outputFrame(bufferToProcess), imageGradient(imageGrad), nbRows(nbRws), nbColumns(nbCols), filterParam_gain(gain){}
         
-        void operator()( const tbb::blocked_range<size_t>& r ) const {
+        virtual void operator()( const Range& r ) const {
             float* offset=outputFrame+nbColumns*nbRows-nbColumns;
             const float* gradOffset= imageGradient+nbColumns*nbRows-nbColumns;
-    	    for (unsigned int IDcolumn=r.begin(); IDcolumn!=r.end(); ++IDcolumn)
+    	    for (int IDcolumn=r.start; IDcolumn!=r.end; ++IDcolumn)
 	    {
 		register float result=0;
 		register float *outputPTR=offset+IDcolumn;
diff --git a/modules/contrib/src/templatebuffer.hpp b/modules/contrib/src/templatebuffer.hpp
index b119b95089..50aedce38a 100644
--- a/modules/contrib/src/templatebuffer.hpp
+++ b/modules/contrib/src/templatebuffer.hpp
@@ -71,15 +71,14 @@
 #include <cmath>
 
 
-//// If TBB is used
+//// If a parallelization method is available then, you should define MAKE_PARALLEL, in the other case, the classical serial code will be used
+#define MAKE_PARALLEL 
 // ==> then include required includes 
-#ifdef HAVE_TBB
-#include "tbb/parallel_for.h"
-#include "tbb/blocked_range.h"
+#ifdef MAKE_PARALLEL
 
 // ==> declare usefull generic tools
 template <class type>
-class Parallel_clipBufferValues
+class Parallel_clipBufferValues: public cv::ParallelLoopBody
 {
 private:
     type *bufferToClip;
@@ -89,9 +88,9 @@ public:
     Parallel_clipBufferValues(type* bufferToProcess, const type min, const type max)
     : bufferToClip(bufferToProcess), minValue(min), maxValue(max){}
 
-    void operator()( const tbb::blocked_range<size_t>& r ) const {
-	register type *inputOutputBufferPTR=bufferToClip+r.begin();
-        for (register unsigned int jf = r.begin(); jf != r.end(); ++jf, ++inputOutputBufferPTR)
+    virtual void operator()( const cv::Range &r ) const {
+	register type *inputOutputBufferPTR=bufferToClip+r.start;
+        for (register int jf = r.start; jf != r.end; ++jf, ++inputOutputBufferPTR)
 	{
 	    if (*inputOutputBufferPTR>maxValue)
 		*inputOutputBufferPTR=maxValue;
@@ -389,8 +388,8 @@ public:
 		std::cout<<"this->min()"<<this->min()<<"minThreshold="<<minThreshold<<"updatedLowValue="<<updatedLowValue<<std::endl;
 		// clipping values outside than the updated thresholds
                 bufferPTR=this->Buffer();
-#ifdef HAVE_TBB // call the TemplateBuffer TBB clipping method
-                tbb::parallel_for(tbb::blocked_range<size_t>(0,this->size()), Parallel_clipBufferValues<type>(bufferPTR, updatedLowValue, updatedHighValue), tbb::auto_partitioner());
+#ifdef MAKE_PARALLEL // call the TemplateBuffer TBB clipping method
+                parallel_for_(tbb::blocked_range<size_t>(0,this->size()), Parallel_clipBufferValues<type>(bufferPTR, updatedLowValue, updatedHighValue));
 #else
 
 		for (unsigned int i=0;i<this->size();++i, ++bufferPTR)