How to adjust Deinterlace filter
The current version of the filter has some methods of filtration; you can choose them on the page "Filter settings".
None - at a choice of this point the filter is disconnected and deinterlacing is not produced.
Only even rows - at a choice of the given method duplication of lines is made, even lines are copied into odd lines without changes. The speed of filtration is very high. Quality of filtration is bad.
Simple interpolation - odd lines are counted on the next even lines by methods of interpolation. The speed of filtration is high. The quality of received image is average. Similar methods are used in the majority of deinterlace filters.
Area based interpolation - only those places on the frame where interlace was found out are interpolated, the other part of the frame remains without changes. The speed of filtration is high. Allows saving those parts of the frame on which there was no movement.
Area based motion estimation - only those parts of the frame on which deinterlace was found out are changed, though restoration is made not by a method of interpolation, but by means of finding vectors of objects displacement. The method is most effective. The speed of filtration is average.
Fields Width and Height allow setting the size of output frame.
Interlace detection threshold - a difference between adjacent pixels when these pixels are considered interlaced. The more this value the algorithm less sensible to the interlace.
Search area - radius in pixels of area where search is fulfiled.
Block distance threshold - the less this value the more similar blocks can inserted into reconstruction field but more blocks will be interpolated.
Replacement threshold - a difference between adjacent pixels when pixels will be interpolated (not be replaced).
In the new version it is added a function of setting the regions for deinterlacing. It is also added an ability of parallel turning out result viewing.
How to adjust:
1. By default the region in which it will be made deinterlace corresponds to the entire frame. Even if region is empty, the entire frame will be deinterlaced.
2. Show output window - to show/hide a preview window. If the tick is established, the preview window will be visible and the filter next after deinterlace one will receive the interlaced frames. The processed frame will be visible in a window of a preview. If the tick is not established, windows of a preview it will not be visible and the filter after deinterlaced filter will receive the processed frame.
3. Processing region – is currently processable region. The region is set by plenty of rectangular areas. There are two types of areas - processing area and not touched area. If the point gets in processing area, on the result frame this point will be processed. If the point gets in not touched area, this point will not be processed. If one point gets in several areas at the same time it will be processed depending on last area type in the list (areas are numbered - see the picture). The area record form has the following appearance:
« [(0,0) - (719,479)] + », where
(0,0) - coordinates of the left top corner of area,
(719, 479) - coordinates of the right bottom corner.
«+» Means that it is areas of processing, «-» - means not touched area.
The sizes of areas should be multiple 8; coordinates of corners of areas should be multiple 8.
4. Add rect - to add area. You can set rectangular area coordinates and set its type manually.
5. Delete rect - to remove area. Deletes marked area from region.
6. Save to file - to save region in a file.
7. Load from file - to load region from a file.
You can set areas directly in a window of a preview. Keeping the left mouse button you will add processing area, keeping the right mouse button you will add not touched area.
The new deinterlacing method - Advanced method is added. It gives the most high-qualitative picture on an output. Low speed of work is a lack of a method. The method has option ME search on interpolated images that allows regulating blocks search process. At the included option blocks search will occur on smoothed picture that leads to more exact blocks finding because there is no noise that can bring a mistake on smoothed picture.