INDEX  F.A.Q.  SEARCH  LATEST POSTS     Rules  Register  Profile  Private messages  Login


Search all forums or this forum: Advanced search
Optimized Conversion between YUV and RGB
Forum Index -> Video -> Programming Printer-friendly version
Reply to topic
Author Message
Dave1024
Member


Joined: 06 Apr 2008
Location: US
Post Posted: May 06, 2008 04:56 Posts Comp View users profile Send private message Reply with quote
Is there any optimized conversion formula available for YUV to RGB and RGB to YUV?
Any formula is available by using shift operations instead of floating point multiplications?

JohnnyMalaria
Member


Joined: 29 May 2006
Location: United States
Post Posted: May 06, 2008 07:11 Posts Comp View users profile Send private message Reply with quote
The usual trick is to do something like this:

Instead of Y = F * X, you multiply both F and X by 256 such that

Y = 256 * 256 * F * X

and then divide by 65536 (256 * 256) to get Y = F * X.

This makes it trivial because if you use 32-bit integers, you either shift right by 24 bits to get your final 8-bit result or just take the high word to get a 16-bit result.

Your input data need to be scaled 0 to 255 (which they usually are) and the coefficients also need to be scaled from 0 to 255.

The next optimization is to use SIMD instructions (MMX/SSE/SSE2 etc) to perform the calculations in parallel.
_________________
John Miller
enosoft - high performance tools for music and video

Home of the Enosoft DV Processor - Free for personal use!

jagabo
Member


Joined: 09 Dec 2005
Location: none
Post Posted: May 06, 2008 07:17 Posts Comp View users profile Send private message Reply with quote
http://fourcc.org/fccyvrgb.php

You can use scaled integers instead of floating point. For example, multiply all values by 256 (<<8) before calculations, then divide the result by 256 (>>8) afterwards. For rounding you can add 128 before dividing by 256.

Is this for some embedded CPU without a FPU? On current x86 processors you may find that using fp SIMD is faster.

Dave1024
Member


Joined: 06 Apr 2008
Location: US
Post Posted: May 07, 2008 00:56 Posts Comp View users profile Send private message Reply with quote
when i use eq: from fourcc and other help files then i got converted image but the picture is blured? Any other factor that the input file depend or not??

jagabo
Member


Joined: 09 Dec 2005
Location: none
Post Posted: May 07, 2008 05:06 Posts Comp View users profile Send private message Reply with quote
How are you getting your YUV video? Are the chroma channels subsampled as YUY2 or YV12 or similar? Maybe you aren't treating the expansion to 4:4:4 correctly? I would do the conversion first in floating point (for simplicity) and worry about speeding it up after the code is working properly.

Dave1024
Member


Joined: 06 Apr 2008
Location: US
Post Posted: May 16, 2008 06:26 Posts Comp View users profile Send private message Reply with quote
Can someone please say about the pixel arrangement when we convert from one format to another...
Is there any relation to clarity of picture and pixel arrangements.
eg: RGB 888 whether entire r value then g and b for frame or r g and b for pixel 1 then pixel 2 ...till frame end

jagabo
Member


Joined: 09 Dec 2005
Location: none
Post Posted: May 16, 2008 07:02 Posts Comp View users profile Send private message Reply with quote
There's no difference in sharpness just because of the storage order of pixels or sub-pixels. There is a loss of sharpness in the color channels when using subsampled YUV (eg, YUY2 or YV12). But order of storage makes no difference.

Dave1024
Member


Joined: 06 Apr 2008
Location: US
Post Posted: May 18, 2008 23:03 Posts Comp View users profile Send private message Reply with quote
can someone tell which are the RGB formats supported by usual 15 inch monitor...
- By use the equation described in www.FourCC.org the required color effect is not obtain. What will i do to get perfect frame?
- What is mean by Alpha channel RGB ...How this is different than other format....

jagabo
Member


Joined: 09 Dec 2005
Location: none
Post Posted: May 19, 2008 07:00 Posts Comp View users profile Send private message Reply with quote
There are many open source video programs. Why don't you download the source code of one of them and compare its conversion algorithm to what you're doing.

An alpha channel usually is 8 bits of additional data added to a 24 bit color image (making it 32 bits per pixel). The extra 8 bits are usually used to indicate transparency.

http://www.hendronix.com/?p=7

Dave1024
Member


Joined: 06 Apr 2008
Location: US
Post Posted: May 19, 2008 23:27 Posts Comp View users profile Send private message Reply with quote
How will we maniculate back buffer for display.
How will we manage the read and write to this back buffer in optimized manner.
plz help me to solve doubts.....

jagabo
Member


Joined: 09 Dec 2005
Location: none
Post Posted: May 20, 2008 07:00 Posts Comp View users profile Send private message Reply with quote
Dave1024 wrote:
How will we maniculate back buffer for display.
How will we manage the read and write to this back buffer in optimized manner.

You read and write to the back buffer exactly the same as the front buffer. When you are done updating the back buffer you call the system function to swap buffers.

In Windows DirectShow you create two Direct Draw surfaces then call DIRECTDRAWSURFACE::Flip() to to swap them for display.

Dave1024
Member


Joined: 06 Apr 2008
Location: US
Post Posted: May 20, 2008 07:08 Posts Comp View users profile Send private message Reply with quote
Direct show and direct X are different package now .Can plz mention direct X API support for back buffer implementation. Is there any other factor relate for getting smooth output with out delay.

Dave1024
Member


Joined: 06 Apr 2008
Location: US
Post Posted: May 26, 2008 01:15 Posts Comp View users profile Send private message Reply with quote
How many minimmum number of buffers need to implement Flipping.
How to load a buffer to the display device.

jagabo
Member


Joined: 09 Dec 2005
Location: none
Post Posted: May 26, 2008 07:58 Posts Comp View users profile Send private message Reply with quote
Dave1024 wrote:
How many minimmum number of buffers need to implement Flipping.

2. Use 3 for smoother results. One being displayed, one fully updated and waiting to be displayed, and one being updated.

Dave1024 wrote:
How to load a buffer to the display device.

http://msdn.microsoft.com/en-us/library/aa139765.aspx

Dave1024
Member


Joined: 06 Apr 2008
Location: US
Post Posted: May 27, 2008 01:40 Posts Comp View users profile Send private message Reply with quote
What is the difference between Texture and Surface?
Is back buffer implemented with both texture and Surface?
What is the importance of locking and unlocking of surface?
Thread Parameters like Mutex,Semaphore,Critical Section have any importance of the video data rendering(Is it use to synchronize read convert and render the video data).

Reply to topic All times are GMT - 6 Hours
Forum Index -> Video -> Programming Page 1 of 1





You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
Jump to:  
Display:   
About   Advertise   Forum Archive   RSS Feeds   Statistics