effects

外国人写的一个很好的嵌入式视频采集源程序。

Effects I coded :
 (see effects.c for more info )
 
 . sobel transformation (works better in 8 bpp) edge detection
 . laplace transformation (works better in 8 bpp) edge detection
 . gamma correction to <value>
 . set threshold to the image (funny effects in RGB24).
 . salt and pepper noise
 . negative image
 . mirroring
 . lowpass (smoothing)
 . highpass (sharpening)
 

 
Well, there is not much to say about them. I dont know if sobel and laplace are ok, I got parts of this code from the net, as I said before, just digging around.
I hope them helps who wants to use videodog for other apps than webcam or security. 
But I couldn`t make every vd user include them or use, so, they are there, but disabled.
To enable is just a matter of uncoment the proper lines  in video.c, grab_one function or insert the proper sequence in grab_loop.
Only the gamma correction effect has an initialization function, the gamma_lookup, to construct a lookup table, and improve the speed.
Send me feedback, more effects code ( I am interested in border detection, contour, crazy transformations, and stuff).



Tips on implementing your effect/processing function



First, check out the template, and fill it with your processing. In this case, Its just looping through the image.
You should check out the vd_video.w for the depth, and apply the right processing rule.
Dont forget to add a prototype to effects.h. The code should stay in effects.c



/* Template for a effect function. */


void myeffect (struct vd_video *vd, ....) {  // The extra parameters are on your  own
         unsigned long int a=0;

         for (a=0; a < vd->grab_size; a=a+vd->w) { // loop through image, in ve->w steps
         	MyImageProcessingAlgorithm(*vd->grab_data + a); // reach each RGB element
		
	}

 }


After that, reach the video.c file and place the call to your effect in the HOOK mark, near a 

/* HOOK for effects */
 
In this point, there will be a complete image in the buffer.
The processing dynamic is the following:

Loop through the buffer, stepping by w bytes each time ( w means the RGB size ).
Do your calculation. 
Done

Things got a little scary when dealing with kernel processing, like sobel or laplace. But its not that difficult, just format your matrix and figure out how to reach the proper elements.



Links:
http://www.cee.hw.ac.uk/hipr/html/hipr_top.html
http://www.ee.uwa.edu.au/~braunl/improv/
http://www.efg2.com/Lab/Library/ImageProcessing/Algorithms.htm


Thats it, Good luck.



Gleicon