Threads in Image processing applications

Question

0.00/5 (No votes)

See more:

Hello everybody,

I'm working in an image processing project. I must use thread because of application's speed. In first thread, I take my image and in second thread I process my image. These must be simultaneously. I think I must use multiple buffer. But how??? Do you have any idea or sample code to do it.

Best regards...

Posted 25-Sep-11 2:02am

AhmetBurak

Add a Solution

Comments

mbue 25-Sep-11 8:16am

Not neccessarily, this depend on the things you want to do. For a single operation (i.e. greyscale) you can break the bitmap in some parts and process every part by a single thread.
i.e. your bitmap has 800 lines and every of your 8 threads processing 100 lines and put the 100 lines to the output bitmap.
Regards.

Richard MacCutchan 25-Sep-11 8:24am

Sorry posted in wrong place.

Richard MacCutchan 25-Sep-11 8:25am

Why does this need multi-threading, can you actually process the image before you finish taking it?

AhmetBurak 25-Sep-11 8:30am

Because of speed. My camera take 50 fps and my aplication must be 50 fps, too. While camera is taking image, I must process my old image.

Sergey Alexandrovich Kryukov 25-Sep-11 15:15pm

Do you realize that one thread should wait for another one anyway? Still you can improve performance the way mbue suggested.
--SA

2 solutions

Add a Solution

Add your solution here

Treat my content as plain text, not as HTML

Preview 0

…

Existing Members

Sign in to your account

...or Join us

Download, Vote, Comment, Publish.

Your Email
Password
Forgot your password?

Your Email
This email is in use. Do you need your password?
Optional Password

I have read and agree to the Terms of Service and Privacy Policy
Please subscribe me to the CodeProject newsletters

When answering a question please:

Read the question carefully.
Understand that English isn't everyone's first language so be lenient of bad spelling and grammar.
If a question is poorly phrased then either ask for clarification, ignore it, or edit the question and fix the problem. Insults are not welcome.
Don't tell someone to read the manual. Chances are they have and don't get it. Provide an answer or move on to the next question.

Let's work to help developers, not make them feel stupid.

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

mbue · Answer 1 · 2011-09-25T05:02:00

I have measured.

C++

  typedef struct
  {
    BITMAPINFO*      pBmp;
    unsigned char*  lpBytes;
    unsigned int    cbBytes;
    unsigned int    firstline;
    unsigned int    linecount;
  } PROCESS_STRUCT;
  static unsigned long FAR PASCAL  __process24(void* p)
  {
    PROCESS_STRUCT*  ps = (PROCESS_STRUCT*)p;
    unsigned int    bpl = ((ps->pBmp->bmiHeader.biWidth*3)+3) & ~3;
    unsigned char*  lpl = ps->lpBytes + bpl * ps->firstline;
    unsigned char*  lpp;
    unsigned int    x,y;
    for(y=0;y<ps->linecount;y++)
    {
      lpp = lpl;
      for(x=0;(int)x<ps->pBmp->bmiHeader.biWidth;x++,lpp+=3)
      {
        lpp[0] = 
        lpp[1] = 
        lpp[2] = (unsigned char)(((unsigned int)lpp[0]+(unsigned int)lpp[1]+(unsigned int)lpp[2])/3);
      }
      lpl += bpl;
    }
    return 0;
  }
  static unsigned long FAR PASCAL  __process32(void* p)
  {
    PROCESS_STRUCT*  ps = (PROCESS_STRUCT*)p;
    unsigned int    bpl = ps->pBmp->bmiHeader.biWidth * 4;
    unsigned char*  lpl = ps->lpBytes + bpl * ps->firstline;
    unsigned char*  lpp;
    unsigned int    x,y;
    for(y=0;y<ps->linecount;y++)
    {
      lpp = lpl;
      for(x=0;(int)x<ps->pBmp->bmiHeader.biWidth;x++,lpp+=4)
      {
        lpp[0] = 
        lpp[1] = 
        lpp[2] = (unsigned char)(((unsigned int)lpp[0]+(unsigned int)lpp[1]+(unsigned int)lpp[2])/3);
      }
      lpl += bpl;
    }
    return 0;
  }

  static void __ProcessBitmap(const unsigned int nthreads,BITMAPINFO* pBmp,unsigned char* lpBytes,unsigned int cbBytes)
  {
    HANDLE              ah[32]; // number of processors == best time
    unsigned int        ih;
    unsigned long        tid;
    PROCESS_STRUCT      pst[32];
    unsigned int        lpp = pBmp->bmiHeader.biHeight / nthreads;

    ASSERT((0<nthreads) && ((sizeof(ah)/sizeof(ah[0]))>=nthreads));

    switch(pBmp->bmiHeader.biBitCount)
    {
      case 24:
        pst[0].firstline = 0;
        for(ih=0;ih<nthreads;ih++)
        {
          if((1+ih)==nthreads) pst[ih].linecount = pBmp->bmiHeader.biHeight-pst[ih].firstline;
          else pst[ih].linecount = min(lpp,pBmp->bmiHeader.biHeight-pst[ih].firstline);
          pst[ih].pBmp    = pBmp;
          pst[ih].lpBytes = lpBytes;
          pst[ih].cbBytes = cbBytes;
          ah[ih] = CreateThread(0,0,__process24,&pst[ih],0,&tid);
          if((1+ih)<nthreads) pst[ih+1].firstline = pst[ih].firstline + pst[ih].linecount;
        }
        WaitForMultipleObjects(ih,ah,1,INFINITE);
      break;
      case 32:
        pst[0].firstline = 0;
        for(ih=0;ih<nthreads;ih++)
        {
          if((1+ih)==nthreads) pst[ih].linecount = pBmp->bmiHeader.biHeight-pst[ih].firstline;
          else pst[ih].linecount = min(lpp,pBmp->bmiHeader.biHeight-pst[ih].firstline);
          pst[ih].pBmp    = pBmp;
          pst[ih].lpBytes = lpBytes;
          pst[ih].cbBytes = cbBytes;
          ah[ih] = CreateThread(0,0,__process32,&pst[ih],0,&tid);
          if((1+ih)<nthreads) pst[ih+1].firstline = pst[ih].firstline + pst[ih].linecount;
        }
        WaitForMultipleObjects(ih,ah,1,INFINITE);
      break;
    }
  }


void ProcessImage(BITMAPINFO* pBmp,unsigned char* lpBytes,unsigned int cbBytes)
{
  // start time measurement
  __ProcessBitmap(_max_threads,pBmp,lpBytes,cbBytes);
  // stop time measurement and add to statistics
}

the results for _max_threads:

C++

  // 1 == 2.64 ms
  // 2 == 1.80 ms
  // 3 == 2.18 ms
  // 4 == 2.00 ms

bitmap is 32 bit 640x480 from the webcam.
on i3 that has 2 physical cores.
Regards.

Albert Holguin · Answer 2 · 2011-09-25T11:08:00

Simple way of doing it... allocate memory on the heap every time there's a new image and pass a pointer to the processing thread. The processing thread then needs to deallocate the memory after its done with it. You can probably have the common access point be a linked list of image pointers and just have the access control be a mutex or something similar.

Threads in Image processing applications

2 solutions

Solution 1

Solution 2

Add your solution here

Preview 0