Convert pixel to YUV image

[Cartouche67]

Hi everybody,

I have an array of pixel in memory, that I obtain like that :

#define PIXEL TDesC8
CWsScreenDevice* screenDev = CCoeEnv::Static()->ScreenDevice();
scr = (PIXEL*) (info().iScreenAddress);

and I wish have this image in an YUV format.

(I already think about : pixel -> bitmap -> YUV but my implementation crachs and I wish have something more optimal.)

Thanks for you help.

[skumar_rao]

If you have the RAW RGB and you want to convert them to yuv then there are no inbuilt APIs in SDK that can do that. But you can use the same Function that you use in Desktop from RGB to YUV in Symbian.C++.

[Cartouche67]

Sorry I don't understand, what you mean "Function that you use in Desktop" ?

[skumar_rao]

I was talking about http://www.fourcc.org/fccyvrgb.php

[skumar_rao]

(I already think about : pixel -> bitmap -> YUV but my implementation crachs and I wish have something more optimal.)

Sorry i over looked this part. It seems you already have a function can you let us know what error you are getting and if you still not able to solve the problem can you show us your function

[Cartouche67]

Thank you for your reply, this is my code :

(
For the RGB to YUV conversion I found the code here : http://wiki.forum.nokia.com/index.ph...YUV_conversion

and for the MDFDevVideo Record I found the code here http://wiki.forum.nokia.com/index.ph...deo_Record_API

)

              #define PIXEL TDesC8

              // Definitions that help access each colour component in source bitmap
              #define   sR ((TInt32)(s[2]))
              #define   sG ((TInt32)(s[1]))
              #define   sB ((TInt32)(s[0]))
 
             #define   KImageWidth     176   // QCIF resolution
             #define   KImageHeigth    144
 
             const TInt KImageNumPixels = KImageWidth * KImageHeigth;


[...]

       CWsScreenDevice* screenDev = CCoeEnv::Static()->ScreenDevice();
	CFbsBitmap* bitmap = new (ELeave) CFbsBitmap;
	User::LeaveIfError(bitmap->Create( TSize(176,144),EColor256 ) ); //TSize(360,640)
	//copy de la capture dans le bitmap
	User::LeaveIfError(screenDev->CopyScreenToBitmap(bitmap));
	
	LFPRINT((_L("avant la conversion")));
	
	/*Debut : code permettant la conversion de bitmap a yuv*/
	
	// Lock source bitmap (CFbsBitmap)
	  bitmap->LockHeap(EFalse);
	  TUint8* s = (TUint8*)bitmap->DataAddress();
	     
	  TInt i = 0;
	  TInt ui = KImageNumPixels;
	  TInt vi = KImageNumPixels + KImageNumPixels/4;
	  
	  // iYuv is an array of TUint8 values, length (KImageNumPixels*3/2)
	  LFPRINT((_L("avant la boucle")));
	    
	  for(TInt j=0; j < KImageHeigth; j++)
	    for(TInt k=0; k < KImageWidth; k++)
	      {
	      //LFPRINT((_L("dans la boucle")));
	      // Y value is generated for each pixel
	      iYuv[i] = (TUint8)( (  66*sR + 129*sG +  25*sB + 128) >> 8 ) + 16;
	      //Mem::Fill(iYuv+i,1,( (  66*sR + 129*sG +  25*sB + 128) >> 8 ) + 16);
	      
	      
	      // U, V values are generated for every other pixel on every other scanline 
	      if(0 == j%2 && 0 == k%2)
	        {
	        iYuv[ui++] = (TUint8)( (-38*sR - 74*sG + 112*sB + 128) >> 8 ) + 128;
	        //Mem::Fill(iYuv+ui,1,( (-38*sR - 74*sG + 112*sB + 128) >> 8 ) + 128);
	        //ui++;
	        iYuv[vi++] = (TUint8)( (112*sR - 94*sG - 18*sB + 128) >> 8 ) + 128;
	        //Mem::Fill(iYuv+vi,1,( (112*sR - 94*sG - 18*sB + 128) >> 8 ) + 128);
	        //vi++;
	        }
	      i++; 
	      s+=EColor256; // Number of bytes representing one pixel in source
	                         // bitmap e.g. if bitmap display mode == EColor16M 
	                         // (24bits/pixel), then iBytesPerPixel == 3
	      }
	  
	  bitmap->UnlockHeap(EFalse);
	  // iYuv now contains the source frame converted to YUV420p format
	  
	  LFPRINT((_L("apres la conversion")));
	
	// Write picture
        LFPRINT((_L("WritePictureL")));
        iPic.iData.iDataFormat = EYuvRawData;
        iPic.iData.iDataSize = TSize(176,144);
        iPicDataPtr.Set(iYuv, 176*144*3/2, 176*144*3/2);
        iPic.iData.iRawData = &iPicDataPtr;
        iPic.iOptions = TVideoPicture::ETimestamp;
        iPic.iTimestamp = (1000000 * iPicNum) / 99;
        TRAPD(err, iDvr->WritePictureL(&iPic));
        if ( err != KErrNone ) {
            LFPRINT((_L("WritePictureL failed with %d"), err));
            TestComplete(err);
        }
        iPicNum++;

I tried to replace iYuv[x]=... with Mem::Fill(iYuv+x... without result.

[skumar_rao]

Can you point out on which line you are crashing.

Note: You can use RFileLogger or Carbide.C++ ODD to debug .

[Cartouche67]

Ok you can find the complete code here http://fr.sourcepod.com/ktdirm14-204

the log file give me this :

-- 14:03:07.380625 - DoTestL...
-- 14:03:07.395625 - NewL()
-- 14:03:07.400625 - FindEncodersL()
-- 14:03:07.520625 - Got 1 encoders
-- 14:03:07.525625 - Get info for encoder 0x10282cfd
-- 14:03:07.705625 - Not accelerated, but haven't seen one so far -- using this
-- 14:03:07.710625 - Selecting encoder 0x10282cfd
-- 14:03:07.750625 - Setting input format
-- 14:03:07.760625 - Setting output format
-- 14:03:07.765625 - Setting buffer options
-- 14:03:07.775625 - SetSourceMemoryL
-- 14:03:07.781250 - Initialize
-- 14:03:07.841250 - MdvroInitializeComplete(0)
-- 14:03:07.846250 - Start
-- 14:03:07.856250 - apres Start
-- 14:03:07.861250 - avant la conversion
-- 14:03:07.871250 - avant la boucle

so there are no logs after the line 304.

And in Carbide.c++ debuger, the last lines are :

 108.830	CVEARMMPEG4EncodeHwDevice::Initialize(), iEncoder [0x4762b648]
   108.830	CVEARMMPEG4EncodeHwDevice::Initialize(), iMaxOutputBufferSize increased to [20480]
   108.830	CVEARMH263EncodeHwDevice::AllocateStructsL(), In
   108.830	CVEARMH263EncodeHwDevice::AllocateStructsL(), iDataArray: [0x4762b908] 
   108.830	CVEARMH263EncodeHwDevice::AllocateStructsL(), [0]: [0x47648814] 
   108.830	CVEARMH263EncodeHwDevice::AllocateStructsL(), [1]: [0x4764d820] 
   108.835	CVEARMH263EncodeHwDevice::AllocateStructsL(), [2]: [0x4765282c] 
   108.835	CVEARMH263EncodeHwDevice::AllocateStructsL(), [3]: [0x47657838] 
   108.835	CVEARMH263EncodeHwDevice::AllocateStructsL(), iOutBufferArray: [0x4762c7b0] 
   108.835	CVEARMH263EncodeHwDevice::AllocateStructsL(), iDataArrayUnc: [0x4762c954] 
   108.835	CVEARMH263EncodeHwDevice::AllocateStructsL(), iOutBuffSize: [0x4762c11c] 
   108.835	CVEARMH263EncodeHwDevice::AllocateStructsL(), iVOPCodingTypePtr: [0x4762c1c4] 
   108.840	CVEARMH263EncodeHwDevice::AllocateStructsL(), iKey[0x47648814] 
   108.840	CVEARMH263EncodeHwDevice::AllocateStructsL(), iKey[0x4764d820] 
   108.840	CVEARMH263EncodeHwDevice::AllocateStructsL(), iKey[0x4765282c] 
   108.840	CVEARMH263EncodeHwDevice::AllocateStructsL(), iKey[0x47657838] 
   108.840	CVEARMH263EncodeHwDevice::AllocateStructsL(), Out
   108.840	CVEARMMPEG4EncodeHwDevice::Initialize(), Initialize status [0]
   108.855	CVEARMMPEG4EncodeHwDevice::Start(), In
   108.855	CVEARMMPEG4EncodeHwDevice::Start(), status[0]
   108.855	CVEARMMPEG4EncodeHwDevice::Start(), Out
   108.880	Thread DevvideoEx::DevvideoEx Panic KERN-EXEC 3

that seems show us that it crashs after the call of Start() line 410, but logs contain the line "apres Start" (that means "after Start" in french), so I don't undertsand where is really the problem.

[skumar_rao]

i suggest you put some more logs to get the exact error Line.

[Cartouche67]

When I added two logs in my program, it crash any more but it doesn't produce a video


I add
LFPRINT((_L("Line 340"))); and LFPRINT((_L("Line 352")));
it appears a lot of time in my log, I'm looking for how put the values of j and k.

[skumar_rao]

I Donot know how you implemented LFPRINT but you can use RFileLogger for easier Logging .

[Cartouche67]

Ok, I managed to write the value of j and k of the loop line 319.

The logs stopped when k=80 and j=6 after the instruction line 324 ...

edit : sorry, I opened the logfile too early, the application had not yet finished.
The values are k=52 and j=54 in the end of the loop.
And the emulator crash when we obtain these values.

[wizard_hu_]

It is always a good idea knowing what kind of data you are working with.
Personally I have absolutely no idea what happens when CopyScreenToBitmap is combined with EColor256 (which is a 256-color => palettized format), but it certainly uses 1 byte per pixel, so accessing "components" (which are actually not there, due to the palettized image) with [0]-[2] does not make any sense.
The

s+=EColor256; // Number of bytes representing one pixel in source

is a "bit" suspicious too. EXy-s are enumerations, the value of EColor256 is actually 4, while it uses 1 byte for representing 1 pixel.
If you are expecting [0]-[2] to mean something, experiment with EColor16MU (perhaps EColor16M).

[Cartouche67]

thank you for your remark wizard, I put EColor16M when I create my bitmap, that mean 1 pixel is coded in 3 bytes, so I put

s+=3;// Number of bytes representing one pixel in source

It crashs any more, and it produce a video in a QCIF format (176*144). Now the problem is that if I change the resolution of capture (I have to register a video in 360*640), it crashs again.
The logs indicates that the call of Start() go wrong, do you have any idee of what can I do ?

edit: I tried with 176*143 (instead 176*144) and it crashs too !

[Cartouche67]

I think it's another specific problem and I'm going to open an second topic.
Thanks for your help !