]> Creatis software - gdcm.git/blobdiff - src/gdcmPixelConvert.cxx
* src/gdcmDocument.cxx : fix bug... test if the fp is opened to use it
[gdcm.git] / src / gdcmPixelConvert.cxx
index 789a600a7c699a905e8d1c9e075f40551857b75a..d3d6b7aa980d70c5d77f1e980cacd173415f7171 100644 (file)
@@ -3,8 +3,8 @@
   Program:   gdcm
   Module:    $RCSfile: gdcmPixelConvert.cxx,v $
   Language:  C++
-  Date:      $Date: 2004/10/08 16:27:20 $
-  Version:   $Revision: 1.3 $
+  Date:      $Date: 2004/11/25 10:24:34 $
+  Version:   $Revision: 1.33 $
                                                                                 
   Copyright (c) CREATIS (Centre de Recherche et d'Applications en Traitement de
   l'Image). All rights reserved. See Doc/License.txt or
                                                                                 
 =========================================================================*/
 
-//////////////////   TEMPORARY NOT
+//////////////////   TEMPORARY NOTE
 // look for "fixMem" and convert that to a member of this class
 // Removing the prefix fixMem and dealing with allocations should do the trick
-#define str2num(str, typeNum) *((typeNum *)(str))
+//
+// grep PIXELCONVERT everywhere and clean up !
 
 #include "gdcmDebug.h"
 #include "gdcmPixelConvert.h"
+#include <fstream>
+#include <stdio.h>
+
+namespace gdcm
+{
+#define str2num(str, typeNum) *((typeNum *)(str))
+
+// For JPEG 2000, body in file gdcmJpeg2000.cxx
+bool gdcm_read_JPEG2000_file (std::ifstream* fp, void* image_buffer);
+
+// For JPEG 8 Bits, body in file gdcmJpeg8.cxx
+bool gdcm_read_JPEG_file8    (std::ifstream* fp, void* image_buffer);
+
+// For JPEG 12 Bits, body in file gdcmJpeg12.cxx
+bool gdcm_read_JPEG_file12   (std::ifstream* fp, void* image_buffer);
+
+// For JPEG 16 Bits, body in file gdcmJpeg16.cxx
+// Beware this is misleading there is no 16bits DCT algorithm, only
+// jpeg lossless compression exist in 16bits.
+bool gdcm_read_JPEG_file16   (std::ifstream* fp, void* image_buffer);
 
 
 //-----------------------------------------------------------------------------
 // Constructor / Destructor
-gdcmPixelConvert::gdcmPixelConvert() 
+PixelConvert::PixelConvert() 
 {
    RGB = 0;
    RGBSize = 0;
-   Uncompressed = 0;
-   UncompressedSize = 0;
+   Decompressed = 0;
+   DecompressedSize = 0;
+   LutRGBA = 0;
+   LutRedData = 0;
+   LutGreenData = 0;
+   LutBlueData =0;
 }
 
-void gdcmPixelConvert::Squeeze() 
+void PixelConvert::Squeeze() 
 {
-   if ( RGB ) {
+   if ( RGB )
+   {
       delete [] RGB;
    } 
-   if ( Uncompressed ) {
-      delete [] Uncompressed;
+   RGB = 0;
+
+   if ( Decompressed )
+   {
+      delete [] Decompressed;
    }
+   Decompressed = 0;
+
+   if ( LutRGBA )
+   {
+      delete [] LutRGBA;
+   }
+   LutRGBA = 0;
 }
 
-gdcmPixelConvert::~gdcmPixelConvert() 
+PixelConvert::~PixelConvert() 
 {
    Squeeze();
 }
 
-void gdcmPixelConvert::AllocateRGB()
+void PixelConvert::AllocateRGB()
 {
   if ( RGB ) {
      delete [] RGB;
   }
-  RGB = new uint8_t[RGBSize];
+  RGB = new uint8_t[ RGBSize ];
 }
 
-void gdcmPixelConvert::AllocateUncompressed()
+void PixelConvert::AllocateDecompressed()
 {
-  if ( Uncompressed ) {
-     delete [] Uncompressed;
+  if ( Decompressed ) {
+     delete [] Decompressed;
   }
-  Uncompressed = new uint8_t[ UncompressedSize ];
+  Decompressed = new uint8_t[ DecompressedSize ];
 }
 
 /**
- * \brief Read from file a 12 bits per pixel image and uncompress it
+ * \brief Read from file a 12 bits per pixel image and decompress it
  *        into a 16 bits per pixel image.
  */
-bool gdcmPixelConvert::ReadAndUncompress12Bits( FILE* filePointer,
-                                                size_t uncompressedSize,
-                                                size_t PixelNumber )
+void PixelConvert::ReadAndDecompress12BitsTo16Bits( std::ifstream* fp )
+               throw ( FormatError )
 {
-   SetUncompressedSize( uncompressedSize );
-   AllocateUncompressed();
+   int nbPixels = XSize * YSize;
+   uint16_t* localDecompres = (uint16_t*)Decompressed;
 
-   uint16_t* pdestination = (uint16_t*)Uncompressed;
-                                                                                
-   for(int p = 0; p < PixelNumber; p += 2 )
+   for( int p = 0; p < nbPixels; p += 2 )
    {
-      // 2 pixels 12bit =     [0xABCDEF]
-      // 2 pixels 16bit = [0x0ABD] + [0x0FCE]
       uint8_t b0, b1, b2;
-      size_t ItemRead;
-      ItemRead = fread( &b0, 1, 1, filePointer);
-      if ( ItemRead != 1 )
+
+      fp->read( (char*)&b0, 1);
+      if ( fp->fail() || fp->eof() )//Fp->gcount() == 1
       {
-         return false;
+         throw FormatError( "PixelConvert::ReadAndDecompress12BitsTo16Bits()",
+                                "Unfound first block" );
       }
-      ItemRead = fread( &b1, 1, 1, filePointer);
-      if ( ItemRead != 1 )
+
+      fp->read( (char*)&b1, 1 );
+      if ( fp->fail() || fp->eof())//Fp->gcount() == 1
       {
-         return false;
+         throw FormatError( "PixelConvert::ReadAndDecompress12BitsTo16Bits()",
+                                "Unfound second block" );
       }
-      ItemRead = fread( &b2, 1, 1, filePointer);
-      if ( ItemRead != 1 )
+
+      fp->read( (char*)&b2, 1 );
+      if ( fp->fail() || fp->eof())//Fp->gcount() == 1
       {
-         return false;
+         throw FormatError( "PixelConvert::ReadAndDecompress12BitsTo16Bits()",
+                                "Unfound second block" );
       }
-                                                                                
-      //Two steps are necessary to please VC++
-      *pdestination++ =  ((b0 >> 4) << 8) + ((b0 & 0x0f) << 4) + (b1 & 0x0f);
-      //                     A                     B                 D
-      *pdestination++ =  ((b2 & 0x0f) << 8) + ((b1 >> 4) << 4) + (b2 >> 4);
-      //                     F                     C                 E
-                                                                                
-      /// \todo JPR Troubles expected on Big-Endian processors ?
-   }
-   return true;
-}
 
-/**
- * \brief Read from file an uncompressed image.
- */
-bool gdcmPixelConvert::ReadUncompressed( FILE* filePointer,
-                                         size_t uncompressedSize,
-                                         size_t expectedSize )
-{
-   if ( expectedSize > uncompressedSize )
-   {
-      dbg.Verbose(0, "gdcmPixelConvert::ReadUncompressed: expectedSize"
-                     "is bigger than it should");
-      return false;
-   }
-   SetUncompressedSize( uncompressedSize );
-   AllocateUncompressed();
-   size_t ItemRead = fread( (void*)Uncompressed, expectedSize, 1, filePointer);
-   if ( ItemRead != 1 )
-   {
-      return false;
-   }
-   return true;
-}
-
-/**
- * \brief  Convert a Gray plane and ( Lut R, Lut G, Lut B ) into an
- *         RGB plane.
- * @return True on success.
- */
-bool gdcmPixelConvert::ConvertGrayAndLutToRGB( uint8_t *lutRGBA )
-
-{
-   /// We assume Uncompressed contains the decompressed gray plane
-   /// and build the RGB image.
-   SetRGBSize( UncompressedSize );
-   AllocateRGB();
+      // Two steps are necessary to please VC++
+      //
+      // 2 pixels 12bit =     [0xABCDEF]
+      // 2 pixels 16bit = [0x0ABD] + [0x0FCE]
+      //                        A                     B                 D
+      *localDecompres++ =  ((b0 >> 4) << 8) + ((b0 & 0x0f) << 4) + (b1 & 0x0f);
+      //                        F                     C                 E
+      *localDecompres++ =  ((b2 & 0x0f) << 8) + ((b1 >> 4) << 4) + (b2 >> 4);
 
-//aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
-//AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
-//COPY HERE THE CODE OF GetImageDataIntoVector
-      
-   /// \todo check that operator new []didn't fail, and sometimes return false
-   return true;
+      /// \todo JPR Troubles expected on Big-Endian processors ?
+   }
 }
 
 /**
  * \brief     Try to deal with RLE 16 Bits. 
  *            We assume the RLE has allready been parsed and loaded in
- *            Uncompressed (through \ref ReadAndUncompressRLE8Bits ).
+ *            Decompressed (through \ref ReadAndDecompressJPEGFile ).
  *            We here need to make 16 Bits Pixels from Low Byte and
  *            High Byte 'Planes'...(for what it may mean)
  * @return    Boolean
  */
-bool gdcmPixelConvert::UncompressRLE16BitsFromRLE8Bits(
-                       int XSize,
-                       int YSize,
-                       int NumberOfFrames,
-                       uint8_t* fixMemUncompressed )
+bool PixelConvert::DecompressRLE16BitsFromRLE8Bits( int NumberOfFrames )
 {
    size_t PixelNumber = XSize * YSize;
-   size_t fixMemUncompressedSize = XSize * YSize * NumberOfFrames;
+   size_t decompressedSize = XSize * YSize * NumberOfFrames;
 
-   // We assumed Uncompressed contains the decoded RLE pixels but as
+   // We assumed Decompressed contains the decoded RLE pixels but as
    // 8 bits per pixel. In order to convert those pixels to 16 bits
-   // per pixel we cannot work in place within Uncompressed and hence
-   // we copy Uncompressed in a safe place, say OldUncompressed.
+   // per pixel we cannot work in place within Decompressed and hence
+   // we copy it in a safe place, say copyDecompressed.
 
-   uint8_t* OldUncompressed = new uint8_t[ fixMemUncompressedSize * 2 ];
-   memmove( OldUncompressed, fixMemUncompressed, fixMemUncompressedSize * 2);
+   uint8_t* copyDecompressed = new uint8_t[ decompressedSize * 2 ];
+   memmove( copyDecompressed, Decompressed, decompressedSize * 2 );
 
-   uint8_t* x = fixMemUncompressed;
-   uint8_t* a = OldUncompressed;
+   uint8_t* x = Decompressed;
+   uint8_t* a = copyDecompressed;
    uint8_t* b = a + PixelNumber;
 
    for ( int i = 0; i < NumberOfFrames; i++ )
    {
-      for ( int j = 0; j < PixelNumber; j++ )
+      for ( unsigned int j = 0; j < PixelNumber; j++ )
       {
-         *(x++) = *(a++);
          *(x++) = *(b++);
+         *(x++) = *(a++);
       }
    }
 
-   delete[] OldUncompressed;
+   delete[] copyDecompressed;
       
    /// \todo check that operator new []didn't fail, and sometimes return false
    return true;
 }
 
 /**
- * \brief Implementation of the RLE decoding algorithm for uncompressing
+ * \brief Implementation of the RLE decoding algorithm for decompressing
  *        a RLE fragment. [refer to PS 3.5-2003, section G.3.2 p 86]
+ * @param subDecompressed Sub region of \ref Decompressed where the de
+ *        decoded fragment should be placed.
+ * @param fragmentSize The length of the binary fragment as found on the disk.
+ * @param decompressedSegmentSize The expected length of the fragment ONCE
+ *        decompressed.
+ * @param fp File Pointer: on entry the position should be the one of
+ *        the fragment to be decoded.
  */
-bool gdcmPixelConvert::ReadAndUncompressRLEFragment( uint8_t* decodedZone,
-                                               long fragmentSize,
-                                               long uncompressedSegmentSize,
-                                               FILE* fp )
+bool PixelConvert::ReadAndDecompressRLEFragment( uint8_t* subDecompressed,
+                                                 long fragmentSize,
+                                                 long decompressedSegmentSize,
+                                                 std::ifstream* fp )
 {
    int8_t count;
    long numberOfOutputBytes = 0;
    long numberOfReadBytes = 0;
-                                                                                
-   while( numberOfOutputBytes < uncompressedSegmentSize )
+
+   while( numberOfOutputBytes < decompressedSegmentSize )
    {
-      fread( &count, 1, 1, fp );
+      fp->read( (char*)&count, 1 );
       numberOfReadBytes += 1;
       if ( count >= 0 )
       // Note: count <= 127 comparison is always true due to limited range
@@ -223,9 +220,9 @@ bool gdcmPixelConvert::ReadAndUncompressRLEFragment( uint8_t* decodedZone,
       //       signed integer of width N is 2^(N-1) - 1, which for int8_t
       //       is 127].
       {
-         fread( decodedZone, count + 1, 1, fp);
-         numberOfReadBytes += count + 1;
-         decodedZone         += count + 1;
+         fp->read( (char*)subDecompressed, count + 1);
+         numberOfReadBytes   += count + 1;
+         subDecompressed     += count + 1;
          numberOfOutputBytes += count + 1;
       }
       else
@@ -233,13 +230,13 @@ bool gdcmPixelConvert::ReadAndUncompressRLEFragment( uint8_t* decodedZone,
          if ( ( count <= -1 ) && ( count >= -127 ) )
          {
             int8_t newByte;
-            fread( &newByte, 1, 1, fp);
+            fp->read( (char*)&newByte, 1);
             numberOfReadBytes += 1;
             for( int i = 0; i < -count + 1; i++ )
             {
-               decodedZone[i] = newByte;
+               subDecompressed[i] = newByte;
             }
-            decodedZone         += -count + 1;
+            subDecompressed     += -count + 1;
             numberOfOutputBytes += -count + 1;
          }
       }
@@ -247,8 +244,8 @@ bool gdcmPixelConvert::ReadAndUncompressRLEFragment( uint8_t* decodedZone,
                                                                                 
       if ( numberOfReadBytes > fragmentSize )
       {
-         dbg.Verbose(0, "gdcmFile::gdcm_read_RLE_fragment: we read more "
-                        "bytes than the segment size.");
+         dbg.Verbose(0, "PixelConvert::ReadAndDecompressRLEFragment: we "
+                        "read more bytes than the segment size.");
          return false;
       }
    }
@@ -257,26 +254,18 @@ bool gdcmPixelConvert::ReadAndUncompressRLEFragment( uint8_t* decodedZone,
 
 /**
  * \brief     Reads from disk the Pixel Data of 'Run Length Encoded'
- *            Dicom encapsulated file and uncompress it.
+ *            Dicom encapsulated file and decompress it.
  * @param     fp already open File Pointer
- * @param     image_buffer destination Address (in caller's memory space)
  *            at which the pixel data should be copied
  * @return    Boolean
  */
-bool gdcmPixelConvert::gdcm_read_RLE_file( void* image_buffer,
-                                   int XSize,
-                                   int YSize,
-                                   int ZSize,
-                                   int BitsAllocated,
-                                   gdcmRLEFramesInfo* RLEInfo,
-                                   FILE* fp )
+bool PixelConvert::ReadAndDecompressRLEFile( std::ifstream* fp )
 {
-   uint8_t* im = (uint8_t*)image_buffer;
-   long uncompressedSegmentSize = XSize * YSize;
-                                                                                
-                                                                                
+   uint8_t* subDecompressed = Decompressed;
+   long decompressedSegmentSize = XSize * YSize;
+
    // Loop on the frame[s]
-   for( gdcmRLEFramesInfo::RLEFrameList::iterator
+   for( RLEFramesInfo::RLEFrameList::iterator
         it  = RLEInfo->Frames.begin();
         it != RLEInfo->Frames.end();
       ++it )
@@ -284,24 +273,878 @@ bool gdcmPixelConvert::gdcm_read_RLE_file( void* image_buffer,
       // Loop on the fragments
       for( unsigned int k = 1; k <= (*it)->NumberFragments; k++ )
       {
-         fseek( fp, (*it)->Offset[k] ,SEEK_SET);
-         (void)gdcmPixelConvert::ReadAndUncompressRLEFragment(
-                                 (uint8_t*) im, (*it)->Length[k],
-                                 uncompressedSegmentSize, fp );
-         im += uncompressedSegmentSize;
+         fp->seekg(  (*it)->Offset[k] , std::ios_base::beg );
+         (void)ReadAndDecompressRLEFragment( subDecompressed,
+                                             (*it)->Length[k],
+                                             decompressedSegmentSize, 
+                                             fp );
+         subDecompressed += decompressedSegmentSize;
       }
    }
-                                                                                
+
    if ( BitsAllocated == 16 )
    {
       // Try to deal with RLE 16 Bits
-      (void)gdcmPixelConvert::UncompressRLE16BitsFromRLE8Bits(
-                                             XSize,
-                                             YSize,
-                                             ZSize,
-                                             (uint8_t*) image_buffer);
+      (void)DecompressRLE16BitsFromRLE8Bits( ZSize );
+   }
+
+   return true;
+}
+
+/**
+ * \brief Swap the bytes, according to \ref SwapCode.
+ */
+void PixelConvert::ConvertSwapZone()
+{
+   unsigned int i;
+
+   if( BitsAllocated == 16 )
+   {
+      uint16_t* im16 = (uint16_t*)Decompressed;
+      switch( SwapCode )
+      {
+         case 0:
+         case 12:
+         case 1234:
+            break;
+         case 21:
+         case 3412:
+         case 2143:
+         case 4321:
+            for( i = 0; i < DecompressedSize / 2; i++ )
+            {
+               im16[i]= (im16[i] >> 8) | (im16[i] << 8 );
+            }
+            break;
+         default:
+            dbg.Verbose( 0, "PixelConvert::ConvertSwapZone: SwapCode value "
+                            "(16 bits) not allowed." );
+      }
+   }
+   else if( BitsAllocated == 32 )
+   {
+      uint32_t s32;
+      uint16_t high;
+      uint16_t low;
+      uint32_t* im32 = (uint32_t*)Decompressed;
+      switch ( SwapCode )
+      {
+         case 0:
+         case 1234:
+            break;
+         case 4321:
+            for( i = 0; i < DecompressedSize / 4; i++ )
+            {
+               low     = im32[i] & 0x0000ffff;  // 4321
+               high    = im32[i] >> 16;
+               high    = ( high >> 8 ) | ( high << 8 );
+               low     = ( low  >> 8 ) | ( low  << 8 );
+               s32     = low;
+               im32[i] = ( s32 << 16 ) | high;
+            }
+            break;
+         case 2143:
+            for( i = 0; i < DecompressedSize / 4; i++ )
+            {
+               low     = im32[i] & 0x0000ffff;   // 2143
+               high    = im32[i] >> 16;
+               high    = ( high >> 8 ) | ( high << 8 );
+               low     = ( low  >> 8 ) | ( low  << 8 );
+               s32     = high;
+               im32[i] = ( s32 << 16 ) | low;
+            }
+            break;
+         case 3412:
+            for( i = 0; i < DecompressedSize / 4; i++ )
+            {
+               low     = im32[i] & 0x0000ffff; // 3412
+               high    = im32[i] >> 16;
+               s32     = low;
+               im32[i] = ( s32 << 16 ) | high;
+            }
+            break;
+         default:
+            dbg.Verbose( 0, "PixelConvert::ConvertSwapZone: SwapCode value "
+                            "(32 bits) not allowed." );
+      }
+   }
+}
+
+/**
+ * \brief Deal with endianity i.e. re-arange bytes inside the integer
+ */
+void PixelConvert::ConvertReorderEndianity()
+{
+   if ( BitsAllocated != 8 )
+   {
+      ConvertSwapZone();
+   }
+
+   // Special kludge in order to deal with xmedcon broken images:
+   if (  ( BitsAllocated == 16 )
+       && ( BitsStored < BitsAllocated )
+       && ( ! PixelSign ) )
+   {
+      int l = (int)( DecompressedSize / ( BitsAllocated / 8 ) );
+      uint16_t *deb = (uint16_t *)Decompressed;
+      for(int i = 0; i<l; i++)
+      {
+         if( *deb == 0xffff )
+         {
+           *deb = 0;
+         }
+         deb++;
+      }
+   }
+}
+
+/**
+ * \brief     Reads from disk the Pixel Data of JPEG Dicom encapsulated
+ &            file and decompress it.
+ * @param     fp File Pointer
+ * @return    Boolean
+ */
+bool PixelConvert::ReadAndDecompressJPEGFile( std::ifstream* fp )
+{
+   uint8_t* localDecompressed = Decompressed;
+   // Loop on the fragment[s]
+   for( JPEGFragmentsInfo::JPEGFragmentsList::iterator
+        it  = JPEGInfo->Fragments.begin();
+        it != JPEGInfo->Fragments.end();
+      ++it )
+   {
+      fp->seekg( (*it)->Offset, std::ios_base::beg);
+
+      if ( IsJPEG2000 )
+      {
+         if ( ! gdcm_read_JPEG2000_file( fp,localDecompressed ) )
+         {
+            return false;
+         }
+      }
+      else if ( BitsStored == 8)
+      {
+         // JPEG Lossy : call to IJG 6b
+         if ( ! gdcm_read_JPEG_file8( fp, localDecompressed ) )
+         {
+            return false;
+         }
+      }
+      else if ( BitsStored <= 12)
+      {
+         // Reading Fragment pixels
+         if ( ! gdcm_read_JPEG_file12 ( fp, localDecompressed ) )
+         {
+            return false;
+         }
+      }
+      else if ( BitsStored <= 16)
+      {
+         // Reading Fragment pixels
+         if ( ! gdcm_read_JPEG_file16 ( fp, localDecompressed ) )
+         {
+            return false;
+         }
+         //assert( IsJPEGLossless );
+      }
+      else
+      {
+         // other JPEG lossy not supported
+         dbg.Error("PixelConvert::ReadAndDecompressJPEGFile: unknown "
+                   "jpeg lossy compression ");
+         return false;
+      }
+
+      // Advance to next free location in Decompressed 
+      // for next fragment decompression (if any)
+      int length = XSize * YSize * SamplesPerPixel;
+      int numberBytes = BitsAllocated / 8;
+
+      localDecompressed += length * numberBytes;
+   }
+   return true;
+}
+
+/**
+ * \brief  Re-arrange the bits within the bytes.
+ * @return Boolean
+ */
+bool PixelConvert::ConvertReArrangeBits() throw ( FormatError )
+{
+   if ( BitsStored != BitsAllocated )
+   {
+      int l = (int)( DecompressedSize / ( BitsAllocated / 8 ) );
+      if ( BitsAllocated == 16 )
+      {
+         uint16_t mask = 0xffff;
+         mask = mask >> ( BitsAllocated - BitsStored );
+         uint16_t* deb = (uint16_t*)Decompressed;
+         for(int i = 0; i<l; i++)
+         {
+            *deb = (*deb >> (BitsStored - HighBitPosition - 1)) & mask;
+            deb++;
+         }
+      }
+      else if ( BitsAllocated == 32 )
+      {
+         uint32_t mask = 0xffffffff;
+         mask = mask >> ( BitsAllocated - BitsStored );
+         uint32_t* deb = (uint32_t*)Decompressed;
+         for(int i = 0; i<l; i++)
+         {
+            *deb = (*deb >> (BitsStored - HighBitPosition - 1)) & mask;
+            deb++;
+         }
+      }
+      else
+      {
+         dbg.Verbose(0, "PixelConvert::ConvertReArrangeBits: weird image");
+         throw FormatError( "PixelConvert::ConvertReArrangeBits()",
+                                "weird image !?" );
+      }
+   }
+   return true;
+}
+
+/**
+ * \brief   Convert (Y plane, cB plane, cR plane) to RGB pixels
+ * \warning Works on all the frames at a time
+ */
+void PixelConvert::ConvertYcBcRPlanesToRGBPixels()
+{
+   uint8_t* localDecompressed = Decompressed;
+   uint8_t* copyDecompressed = new uint8_t[ DecompressedSize ];
+   memmove( copyDecompressed, localDecompressed, DecompressedSize );
+
+   // to see the tricks about YBR_FULL, YBR_FULL_422,
+   // YBR_PARTIAL_422, YBR_ICT, YBR_RCT have a look at :
+   // ftp://medical.nema.org/medical/dicom/final/sup61_ft.pdf
+   // and be *very* affraid
+   //
+   int l        = XSize * YSize;
+   int nbFrames = ZSize;
+
+   uint8_t* a = copyDecompressed;
+   uint8_t* b = copyDecompressed + l;
+   uint8_t* c = copyDecompressed + l + l;
+   double R, G, B;
+
+   /// \todo : Replace by the 'well known' integer computation
+   ///         counterpart. Refer to
+   ///            http://lestourtereaux.free.fr/papers/data/yuvrgb.pdf
+   ///         for code optimisation.
+
+   for ( int i = 0; i < nbFrames; i++ )
+   {
+      for ( int j = 0; j < l; j++ )
+      {
+         R = 1.164 *(*a-16) + 1.596 *(*c -128) + 0.5;
+         G = 1.164 *(*a-16) - 0.813 *(*c -128) - 0.392 *(*b -128) + 0.5;
+         B = 1.164 *(*a-16) + 2.017 *(*b -128) + 0.5;
+
+         if (R < 0.0)   R = 0.0;
+         if (G < 0.0)   G = 0.0;
+         if (B < 0.0)   B = 0.0;
+         if (R > 255.0) R = 255.0;
+         if (G > 255.0) G = 255.0;
+         if (B > 255.0) B = 255.0;
+
+         *(localDecompressed++) = (uint8_t)R;
+         *(localDecompressed++) = (uint8_t)G;
+         *(localDecompressed++) = (uint8_t)B;
+         a++;
+         b++;
+         c++;
+      }
+   }
+   delete[] copyDecompressed;
+}
+
+/**
+ * \brief   Convert (Red plane, Green plane, Blue plane) to RGB pixels
+ * \warning Works on all the frames at a time
+ */
+void PixelConvert::ConvertRGBPlanesToRGBPixels()
+{
+   uint8_t* localDecompressed = Decompressed;
+   uint8_t* copyDecompressed = new uint8_t[ DecompressedSize ];
+   memmove( copyDecompressed, localDecompressed, DecompressedSize );
+
+   int l = XSize * YSize * ZSize;
+
+   uint8_t* a = copyDecompressed;
+   uint8_t* b = copyDecompressed + l;
+   uint8_t* c = copyDecompressed + l + l;
+
+   for (int j = 0; j < l; j++)
+   {
+      *(localDecompressed++) = *(a++);
+      *(localDecompressed++) = *(b++);
+      *(localDecompressed++) = *(c++);
+   }
+   delete[] copyDecompressed;
+}
+
+bool PixelConvert::ReadAndDecompressPixelData( std::ifstream* fp )
+{
+   // ComputeDecompressedAndRGBSizes is already made by 
+   // ::GrabInformationsFromHeader. So, the structure sizes are
+   // correct
+   Squeeze();
+
+   //////////////////////////////////////////////////
+   //// First stage: get our hands on the Pixel Data.
+   if ( !fp )
+   {
+      dbg.Verbose( 0, "PixelConvert::ReadAndDecompressPixelData: "
+                      "unavailable file pointer." );
+      return false;
+   }
+
+   fp->seekg( PixelOffset, std::ios_base::beg );
+   if( fp->fail() || fp->eof()) //Fp->gcount() == 1
+   {
+      dbg.Verbose( 0, "PixelConvert::ReadAndDecompressPixelData: "
+                      "unable to find PixelOffset in file." );
+      return false;
+   }
+
+   AllocateDecompressed();
+
+   //////////////////////////////////////////////////
+   //// Second stage: read from disk dans decompress.
+   if ( BitsAllocated == 12 )
+   {
+      ReadAndDecompress12BitsTo16Bits( fp);
+   }
+   else if ( IsDecompressed )
+   {
+      // This problem can be found when some obvious informations are found
+      // after the field containing the image datas. In this case, these
+      // bad datas are added to the size of the image (in the PixelDataLength
+      // variable). But DecompressedSize is the right size of the image !
+      if( PixelDataLength != DecompressedSize)
+      {
+         dbg.Verbose( 0, "PixelConvert::ReadAndDecompressPixelData: "
+                      "Mismatch between PixelConvert and DecompressedSize." );
+      }
+      if( PixelDataLength > DecompressedSize)
+      {
+         fp->read( (char*)Decompressed, DecompressedSize);
+      }
+      else
+      {
+         fp->read( (char*)Decompressed, PixelDataLength);
+      }
+
+      if ( fp->fail() || fp->eof())//Fp->gcount() == 1
+      {
+         dbg.Verbose( 0, "PixelConvert::ReadAndDecompressPixelData: "
+                         "reading of decompressed pixel data failed." );
+         return false;
+      }
+   } 
+   else if ( IsRLELossless )
+   {
+      if ( ! ReadAndDecompressRLEFile( fp ) )
+      {
+         dbg.Verbose( 0, "PixelConvert::ReadAndDecompressPixelData: "
+                         "RLE decompressor failed." );
+         return false;
+      }
+   }
+   else
+   {
+      // Default case concerns JPEG family
+      if ( ! ReadAndDecompressJPEGFile( fp ) )
+      {
+         dbg.Verbose( 0, "PixelConvert::ReadAndDecompressPixelData: "
+                         "JPEG decompressor failed." );
+         return false;
+      }
+   }
+
+   ////////////////////////////////////////////
+   //// Third stage: twigle the bytes and bits.
+   ConvertReorderEndianity();
+   ConvertReArrangeBits();
+   ConvertHandleColor();
+
+   return true;
+}
+
+void PixelConvert::ConvertHandleColor()
+{
+   //////////////////////////////////
+   // Deal with the color decoding i.e. handle:
+   //   - R, G, B planes (as opposed to RGB pixels)
+   //   - YBR (various) encodings.
+   //   - LUT[s] (or "PALETTE COLOR").
+   //
+   // The classification in the color decoding schema is based on the blending
+   // of two Dicom tags values:
+   // * "Photometric Interpretation" for which we have the cases:
+   //  - [Photo A] MONOCHROME[1|2] pictures,
+   //  - [Photo B] RGB or YBR_FULL_422 (which acts as RGB),
+   //  - [Photo C] YBR_* (with the above exception of YBR_FULL_422)
+   //  - [Photo D] "PALETTE COLOR" which indicates the presence of LUT[s].
+   // * "Planar Configuration" for which we have the cases:
+   //  - [Planar 0] 0 then Pixels are already RGB
+   //  - [Planar 1] 1 then we have 3 planes : R, G, B,
+   //  - [Planar 2] 2 then we have 1 gray Plane and 3 LUTs
+   //
+   // Now in theory, one could expect some coherence when blending the above
+   // cases. For example we should not encounter files belonging at the
+   // time to case [Planar 0] and case [Photo D].
+   // Alas, this was only theory ! Because in practice some odd (read ill
+   // formated Dicom) files (e.g. gdcmData/US-PAL-8-10x-echo.dcm) we encounter:
+   //     - "Planar Configuration" = 0,
+   //     - "Photometric Interpretation" = "PALETTE COLOR".
+   // Hence gdcm shall use the folowing "heuristic" in order to be tolerant
+   // towards Dicom-non-conformance files:
+   //   << whatever the "Planar Configuration" value might be, a
+   //      "Photometric Interpretation" set to "PALETTE COLOR" forces
+   //      a LUT intervention >>
+   //
+   // Now we are left with the following handling of the cases:
+   // - [Planar 0] OR  [Photo A] no color decoding (since respectively
+   //       Pixels are already RGB and monochrome pictures have no color :),
+   // - [Planar 1] AND [Photo B] handled with ConvertRGBPlanesToRGBPixels()
+   // - [Planar 1] AND [Photo C] handled with ConvertYcBcRPlanesToRGBPixels()
+   // - [Planar 2] OR  [Photo D] requires LUT intervention.
+
+   if ( ! IsDecompressedRGB() )
+   {
+      // [Planar 2] OR  [Photo D]: LUT intervention done outside
+      return;
+   }
+                                                                                
+   if ( PlanarConfiguration == 1 )
+   {
+      if ( IsYBRFull )
+      {
+         // [Planar 1] AND [Photo C] (remember YBR_FULL_422 acts as RGB)
+         ConvertYcBcRPlanesToRGBPixels();
+      }
+      else
+      {
+         // [Planar 1] AND [Photo C]
+         ConvertRGBPlanesToRGBPixels();
+      }
+      return;
    }
                                                                                 
+   // When planarConf is 0, and RLELossless (forbidden by Dicom norm)
+   // pixels need to be RGB-fied anyway
+   if (IsRLELossless)
+   {
+     ConvertRGBPlanesToRGBPixels();
+   }
+   // In *normal *case, when planarConf is 0, pixels are already in RGB
+}
+
+/**
+ * \brief Predicate to know wether the image[s] (once decompressed) is RGB.
+ * \note See comments of \ref ConvertHandleColor
+ */
+bool PixelConvert::IsDecompressedRGB()
+{
+   if (   IsMonochrome
+       || PlanarConfiguration == 2
+       || IsPaletteColor )
+   {
+      return false;
+   }
    return true;
 }
 
+void PixelConvert::ComputeDecompressedAndRGBSizes()
+{
+   int bitsAllocated = BitsAllocated;
+   // Number of "Bits Allocated" is fixed to 16 when it's 12, since
+   // in this case we will expand the image to 16 bits (see
+   //    \ref ReadAndDecompress12BitsTo16Bits() )
+   if (  BitsAllocated == 12 )
+   {
+      bitsAllocated = 16;
+   }
+                                                                                
+   DecompressedSize =  XSize * YSize * ZSize
+                     * ( bitsAllocated / 8 )
+                     * SamplesPerPixel;
+   if ( HasLUT )
+   {
+      RGBSize = 3 * DecompressedSize;
+   }
+   else
+   {
+      RGBSize = DecompressedSize;
+   }
+}
+
+void PixelConvert::GrabInformationsFromHeader( Header* header )
+{
+   // Just in case some access to a Header element requires disk access.
+   // Note: gdcmDocument::Fp is leaved open after OpenFile.
+   std::ifstream* fp = header->OpenFile();
+   // Number of Bits Allocated for storing a Pixel is defaulted to 16
+   // when absent from the header.
+   BitsAllocated = header->GetBitsAllocated();
+   if ( BitsAllocated == 0 )
+   {
+      BitsAllocated = 16;
+   }
+
+   // Number of "Bits Stored" defaulted to number of "Bits Allocated"
+   // when absent from the header.
+   BitsStored = header->GetBitsStored();
+   if ( BitsStored == 0 )
+   {
+      BitsStored = BitsAllocated;
+   }
+
+   // High Bit Position
+   HighBitPosition = header->GetHighBitPosition();
+   if ( HighBitPosition == 0 )
+   {
+      HighBitPosition = BitsAllocated - 1;
+   }
+
+   XSize = header->GetXSize();
+   YSize = header->GetYSize();
+   ZSize = header->GetZSize();
+   SamplesPerPixel = header->GetSamplesPerPixel();
+   PixelSize = header->GetPixelSize();
+   PixelSign = header->IsSignedPixelData();
+   SwapCode  = header->GetSwapCode();
+   TransferSyntaxType ts = header->GetTransferSyntax();
+   IsDecompressed =
+        ( ! header->IsDicomV3() )
+     || ts == ImplicitVRLittleEndian
+     || ts == ImplicitVRLittleEndianDLXGE
+     || ts == ExplicitVRLittleEndian
+     || ts == ExplicitVRBigEndian
+     || ts == DeflatedExplicitVRLittleEndian;
+   IsJPEG2000     = header->IsJPEG2000();
+   IsJPEGLossless = header->IsJPEGLossless();
+   IsRLELossless  =  ( ts == RLELossless );
+   PixelOffset     = header->GetPixelOffset();
+   PixelDataLength = header->GetPixelAreaLength();
+   RLEInfo  = header->GetRLEInfo();
+   JPEGInfo = header->GetJPEGInfo();
+                                                                             
+   PlanarConfiguration = header->GetPlanarConfiguration();
+   IsMonochrome = header->IsMonochrome();
+   IsPaletteColor = header->IsPaletteColor();
+   IsYBRFull = header->IsYBRFull();
+
+   /////////////////////////////////////////////////////////////////
+   // LUT section:
+   HasLUT = header->HasLUT();
+   if ( HasLUT )
+   {
+      LutRedDescriptor   = header->GetEntryByNumber( 0x0028, 0x1101 );
+      LutGreenDescriptor = header->GetEntryByNumber( 0x0028, 0x1102 );
+      LutBlueDescriptor  = header->GetEntryByNumber( 0x0028, 0x1103 );
+   
+      // Depending on the value of Document::MAX_SIZE_LOAD_ELEMENT_VALUE
+      // [ refer to invocation of Document::SetMaxSizeLoadEntry() in
+      // Document::Document() ], the loading of the value (content) of a
+      // [Bin|Val]Entry occurence migth have been hindered (read simply NOT
+      // loaded). Hence, we first try to obtain the LUTs data from the header
+      // and when this fails we read the LUTs data directely from disk.
+      /// \todo Reading a [Bin|Val]Entry directly from disk is a kludge.
+      ///       We should NOT bypass the [Bin|Val]Entry class. Instead
+      ///       an access to an UNLOADED content of a [Bin|Val]Entry occurence
+      ///       (e.g. BinEntry::GetBinArea()) should force disk access from
+      ///       within the [Bin|Val]Entry class itself. The only problem
+      ///       is that the [Bin|Val]Entry is unaware of the FILE* is was
+      ///       parsed from. Fix that. FIXME.
+   
+      ////// Red round:
+      LutRedData = (uint8_t*)header->GetEntryBinAreaByNumber( 0x0028, 0x1201 );
+      if ( ! LutRedData )
+      {
+         // Read the Lut Data from disk
+         DocEntry* lutRedDataEntry = header->GetDocEntryByNumber( 0x0028,
+                                                                  0x1201 );
+         LutRedData = new uint8_t[ lutRedDataEntry->GetLength() ];
+         fp->seekg(  lutRedDataEntry->GetOffset() ,std::ios_base::beg );
+         fp->read( (char*)LutRedData, (size_t)lutRedDataEntry->GetLength());
+         if ( fp->fail() || fp->eof())//Fp->gcount() == 1
+         {
+            dbg.Verbose(0, "PixelConvert::GrabInformationsFromHeader: "
+                            "unable to read red LUT data" );
+         }
+      }
+
+      ////// Green round:
+      LutGreenData = (uint8_t*)header->GetEntryBinAreaByNumber(0x0028, 0x1202 );
+      if ( ! LutGreenData)
+      {
+         // Read the Lut Data from disk
+         DocEntry* lutGreenDataEntry = header->GetDocEntryByNumber( 0x0028,
+                                                                    0x1202 );
+         LutGreenData = new uint8_t[ lutGreenDataEntry->GetLength() ];
+         fp->seekg( lutGreenDataEntry->GetOffset() , std::ios_base::beg );
+         fp->read( (char*)LutGreenData, (size_t)lutGreenDataEntry->GetLength() );
+         if ( fp->fail() || fp->eof())//Fp->gcount() == 1
+         {
+            dbg.Verbose(0, "PixelConvert::GrabInformationsFromHeader: "
+                           "unable to read green LUT data" );
+         }
+      }
+
+      ////// Blue round:
+      LutBlueData = (uint8_t*)header->GetEntryBinAreaByNumber( 0x0028, 0x1203 );
+      if ( ! LutBlueData )
+      {
+         // Read the Lut Data from disk
+         DocEntry* lutBlueDataEntry  = header->GetDocEntryByNumber( 0x0028,
+                                                                    0x1203 );
+         LutBlueData = new uint8_t[ lutBlueDataEntry->GetLength() ];
+         fp->seekg(  lutBlueDataEntry->GetOffset() , std::ios_base::beg );
+         fp->read( (char*)LutBlueData, (size_t)lutBlueDataEntry->GetLength() );
+         if ( fp->fail() || fp->eof())//Fp->gcount() == 1
+         {
+            dbg.Verbose(0, "PixelConvert::GrabInformationsFromHeader: "
+                           "unable to read blue LUT data" );
+         }
+      }
+   }
+
+   ComputeDecompressedAndRGBSizes();
+
+   if(fp) 
+   {
+      header->CloseFile();
+   }
+}
+
+/**
+ * \brief Build Red/Green/Blue/Alpha LUT from Header
+ *         when (0028,0004),Photometric Interpretation = [PALETTE COLOR ]
+ *          and (0028,1101),(0028,1102),(0028,1102)
+ *            - xxx Palette Color Lookup Table Descriptor - are found
+ *          and (0028,1201),(0028,1202),(0028,1202)
+ *            - xxx Palette Color Lookup Table Data - are found
+ * \warning does NOT deal with :
+ *   0028 1100 Gray Lookup Table Descriptor (Retired)
+ *   0028 1221 Segmented Red Palette Color Lookup Table Data
+ *   0028 1222 Segmented Green Palette Color Lookup Table Data
+ *   0028 1223 Segmented Blue Palette Color Lookup Table Data
+ *   no known Dicom reader deals with them :-(
+ * @return a RGBA Lookup Table
+ */
+void PixelConvert::BuildLUTRGBA()
+{
+   if ( LutRGBA )
+   {
+      return;
+   }
+   // Not so easy : see
+   // http://www.barre.nom.fr/medical/dicom2/limitations.html#Color%20Lookup%20Tables
+                                                                                
+   if ( ! IsPaletteColor )
+   {
+      return;
+   }
+                                                                                
+   if (   LutRedDescriptor   == GDCM_UNFOUND
+       || LutGreenDescriptor == GDCM_UNFOUND
+       || LutBlueDescriptor  == GDCM_UNFOUND )
+   {
+      return;
+   }
+
+   ////////////////////////////////////////////
+   // Extract the info from the LUT descriptors
+   int lengthR;   // Red LUT length in Bytes
+   int debR;      // Subscript of the first Lut Value
+   int nbitsR;    // Lut item size (in Bits)
+   int nbRead = sscanf( LutRedDescriptor.c_str(),
+                        "%d\\%d\\%d",
+                        &lengthR, &debR, &nbitsR );
+   if( nbRead != 3 )
+   {
+      dbg.Verbose(0, "PixelConvert::BuildLUTRGBA: wrong red LUT descriptor");
+   }
+                                                                                
+   int lengthG;  // Green LUT length in Bytes
+   int debG;     // Subscript of the first Lut Value
+   int nbitsG;   // Lut item size (in Bits)
+   nbRead = sscanf( LutGreenDescriptor.c_str(),
+                    "%d\\%d\\%d",
+                    &lengthG, &debG, &nbitsG );
+   if( nbRead != 3 )
+   {
+      dbg.Verbose(0, "PixelConvert::BuildLUTRGBA: wrong green LUT descriptor");
+   }
+                                                                                
+   int lengthB;  // Blue LUT length in Bytes
+   int debB;     // Subscript of the first Lut Value
+   int nbitsB;   // Lut item size (in Bits)
+   nbRead = sscanf( LutRedDescriptor.c_str(),
+                    "%d\\%d\\%d",
+                    &lengthB, &debB, &nbitsB );
+   if( nbRead != 3 )
+   {
+      dbg.Verbose(0, "PixelConvert::BuildLUTRGBA: wrong blue LUT descriptor");
+   }
+                                                                                
+   ////////////////////////////////////////////////////////
+   if ( ( ! LutRedData ) || ( ! LutGreenData ) || ( ! LutBlueData ) )
+   {
+      return;
+   }
+
+   ////////////////////////////////////////////////
+   // forge the 4 * 8 Bits Red/Green/Blue/Alpha LUT
+   LutRGBA = new uint8_t[ 1024 ]; // 256 * 4 (R, G, B, Alpha)
+   if ( !LutRGBA )
+   {
+      return;
+   }
+   memset( LutRGBA, 0, 1024 );
+                                                                                
+   int mult;
+   if ( ( nbitsR == 16 ) && ( BitsAllocated == 8 ) )
+   {
+      // when LUT item size is different than pixel size
+      mult = 2; // high byte must be = low byte
+   }
+   else
+   {
+      // See PS 3.3-2003 C.11.1.1.2 p 619
+      mult = 1;
+   }
+                                                                                
+   // if we get a black image, let's just remove the '+1'
+   // from 'i*mult+1' and check again
+   // if it works, we shall have to check the 3 Palettes
+   // to see which byte is ==0 (first one, or second one)
+   // and fix the code
+   // We give up the checking to avoid some (useless ?)overhead
+   // (optimistic asumption)
+   int i;
+   uint8_t* a = LutRGBA + 0;
+   for( i=0; i < lengthR; ++i )
+   {
+      *a = LutRedData[i*mult+1];
+      a += 4;
+   }
+                                                                                
+   a = LutRGBA + 1;
+   for( i=0; i < lengthG; ++i)
+   {
+      *a = LutGreenData[i*mult+1];
+      a += 4;
+   }
+                                                                                
+   a = LutRGBA + 2;
+   for(i=0; i < lengthB; ++i)
+   {
+      *a = LutBlueData[i*mult+1];
+      a += 4;
+   }
+                                                                                
+   a = LutRGBA + 3;
+   for(i=0; i < 256; ++i)
+   {
+      *a = 1; // Alpha component
+      a += 4;
+   }
+}
+
+/**
+ * \brief Build the RGB image from the Decompressed imagage and the LUTs.
+ */
+bool PixelConvert::BuildRGBImage()
+{
+   if ( RGB )
+   {
+      // The job is already done.
+      return true;
+   }
+
+   if ( ! Decompressed )
+   {
+      // The job can't be done
+      return false;
+   }
+
+   BuildLUTRGBA();
+   if ( ! LutRGBA )
+   {
+      // The job can't be done
+      return false;
+   }
+                                                                                
+   // Build RGB Pixels
+   AllocateRGB();
+   uint8_t* localRGB = RGB;
+   for (size_t i = 0; i < DecompressedSize; ++i )
+   {
+      int j  = Decompressed[i] * 4;
+      *localRGB++ = LutRGBA[j];
+      *localRGB++ = LutRGBA[j+1];
+      *localRGB++ = LutRGBA[j+2];
+   }
+   return true;
+}
+
+/**
+ * \brief        Print self.
+ * @param indent Indentation string to be prepended during printing.
+ * @param os     Stream to print to.
+ */
+void PixelConvert::Print( std::string indent, std::ostream &os )
+{
+   os << indent
+      << "--- Pixel information -------------------------"
+      << std::endl;
+   os << indent
+      << "Pixel Data: offset " << PixelOffset
+      << " x" << std::hex << PixelOffset << std::dec
+      << "   length " << PixelDataLength
+      << " x" << std::hex << PixelDataLength << std::dec
+      << std::endl;
+
+   if ( IsRLELossless )
+   {
+      if ( RLEInfo )
+      {
+         RLEInfo->Print( indent, os );
+      }
+      else
+      {
+         dbg.Verbose(0, "PixelConvert::Print: set as RLE file "
+                        "but NO RLEinfo present.");
+      }
+   }
+
+   if ( IsJPEG2000 || IsJPEGLossless )
+   {
+      if ( JPEGInfo )
+      {
+         JPEGInfo->Print( indent, os );
+      }
+      else
+      {
+         dbg.Verbose(0, "PixelConvert::Print: set as JPEG file "
+                        "but NO JPEGinfo present.");
+      }
+   }
+}
+
+} // end namespace gdcm
+
+// NOTES on File internal calls
+// User
+// ---> GetImageData
+//     ---> GetImageDataIntoVector
+//        |---> GetImageDataIntoVectorRaw
+//        | lut intervention
+// User
+// ---> GetImageDataRaw
+//     ---> GetImageDataIntoVectorRaw