diff --git a/Modules/Filtering/Statistics/include/otbListSampleGenerator.txx b/Modules/Filtering/Statistics/include/otbListSampleGenerator.txx
index cc980403131a91298cef2727853f6cd75d9ba124..85fcebc15ecc6fd7d35979d38ec15fd41d19048c 100644
--- a/Modules/Filtering/Statistics/include/otbListSampleGenerator.txx
+++ b/Modules/Filtering/Statistics/include/otbListSampleGenerator.txx
@@ -400,23 +400,30 @@ ListSampleGenerator<TImage, TVectorData>
        itmap != m_ClassesSize.end();
        ++itmap)
     {
-    m_ClassesProbTraining[itmap->first] = minSizeTraining / itmap->second;
-    m_ClassesProbValidation[itmap->first] = minSizeValidation / itmap->second;
-    if(!m_BoundByMin)
+    if (m_BoundByMin)
+      {
+      m_ClassesProbTraining[itmap->first] = minSizeTraining / itmap->second;
+      m_ClassesProbValidation[itmap->first] = minSizeValidation / itmap->second;
+      }
+    else
       {
       long int maxSizeT = (itmap->second)*(1.0 - m_ValidationTrainingProportion);
       long int maxSizeV = (itmap->second)*m_ValidationTrainingProportion;
-      maxSizeT = (m_MaxTrainingSize == -1)?maxSizeT:m_MaxTrainingSize;
-      maxSizeV = (m_MaxValidationSize == -1)?maxSizeV:m_MaxValidationSize;
-    
-      //not enough samples to respect the bounds
-      if(maxSizeT+maxSizeV > itmap->second)
+
+      // Check if max sizes respect the maximum bounds
+      double correctionRatioTrain = 1.0;
+      if((m_MaxTrainingSize > -1) && (m_MaxTrainingSize < maxSizeT))
+        {
+        correctionRatioTrain = (double)(m_MaxTrainingSize) / (double)(maxSizeT);
+        }
+      double correctionRatioValid = 1.0;
+      if((m_MaxValidationSize > -1) && (m_MaxValidationSize < maxSizeV))
         {
-        maxSizeT = (itmap->second)*(1.0 - m_ValidationTrainingProportion);
-        maxSizeV = (itmap->second)*m_ValidationTrainingProportion;
+        correctionRatioValid = (double)(m_MaxValidationSize) / (double)(maxSizeV);
         }
-      m_ClassesProbTraining[itmap->first] = maxSizeT/(itmap->second);
-      m_ClassesProbValidation[itmap->first] = maxSizeV/(itmap->second);
+      double correctionRatio = std::min(correctionRatioTrain,correctionRatioValid);
+      m_ClassesProbTraining[itmap->first] = correctionRatio*(1.0 - m_ValidationTrainingProportion);
+      m_ClassesProbValidation[itmap->first] = correctionRatio*m_ValidationTrainingProportion;
       }
     }
 }