Alparysoft R&DInnovative Multimedia Solutions
Video processing, filtration, compression, image recognition and video surveillance security systems
Home Company Products Services Downloads Purchase Subscription Site Map
 
Road Surface Monitor
 
Home / Articles / Comparisons of the filters intended for removing interlace effect

Comparisons of the filters intended for removing interlace effect

Comparisons of the filters intended for removing interlace effect

Our task was to compare the quality of removing interlace effect from video images. For this purpose some filters, eliminating interlace (Deinterlace filter) were chosen:

  1. Alparysoft Deinterlace Filter (ver. 1.0 beta, build 082)
  2. Smart Deinterlacing filter by Donald Graft (ver. 2.7 beta)
  3. Virtual Dub internal deinterlace filter (ver. 1.5.9)

As far as on real interlaced films the quality of a filtration can be estimated only subjectively (in a visual way), we have chosen following method:

  1. One got out any video fragment M1 of high quality without interlace effect.
  2. Our filter of imposing artificial interlace was applied to the chosen video. The fragment M2 occurred.
  3. Filters were put in turn to the fragment M2. The fragment M3 occurred.
  4. Compare the original video fragment M1 with M3.

Artificial interlace was got out by means of two frames overlapping in a line, the distance between them was determined a priori given constant N. The following fragment of a code shows an example of this method realization.

// frame dimentions
extern int width, height;
// pointer to current frame and previous.
extern unsigned char *pframe_cur, *pframe_prev;
int i;
for(i=0; i < height; i+=2)
 memcpy(pframe_cur + i * width,pframe_prev + i * width, width);

Here pframe_cur and pframe_prev are pointers to frame's buffers with width*height*bits_per_pixel size in bytes.

The above described technique of Artificial interlace reception is realized by us as the DirectShow filter. In case you are interested in it, we may send it to you on your inquiry.

For two images comparison the mean square error was used:

Where P and Q - the frames shown as matrixes by dimension w on h; values is a pixel's brightness.

The comparative characteristics of the filters removing interlace are given here


fig. 1. Mean square error between original and filted image.



If you can offer an another method for comparison deinterlacing quality, write to us. We shell try to publish comparison results here.