|
Adaptive Per-Pixel De-Interlacing
Digital displays, like HDTVs, DVI displays, and CRT monitors, ultimately require a progressive scan signal to operate properly. If they receive an interlaced signal, it must be converted to a progressive scan signal before it can be displayed on the digital monitor.
Interlaced video images are made up of fields. Each field contains half of the number of lines needed to make up one frame of video. A field consists of either the even or odd lines of the original image. During interlace display scanning the screen is refreshed in two top-to-bottom passes such that the lines scanned in one pass are positioned between the lines drawn in the previous pass. Digital displays use progressive scanning, which means that they render all lines in a single top-to-bottom pass, which requires twice as much data per pass as interlaced scanning. Therefore, the video data format must be converted from one that is compatible with interlaced fields to one that provides progressive frames prior to rendering on a digital display. The process of translating received interlaced video signals into a progressive scan format for output and display on any digital display is called de-interlacing.
|
|
|
Interlaced Format versus De-Interlaced Format |
There are two straightforward and basic methods of de-interlacing an interlaced video image for display on a progressive scan monitor: “ Bob ” and “ Weave .”
" Bob " (intra-field spatial interpolation) is the process of completely discarding one of the fields in an interlaced video data. Either all the odd or all the even fields are discarded. Using the fields that remain, interpolation is performed between the lines to generate an entire non-interlaced frame for the progressive scan monitor. This method includes no reference to the discarded fields.
" Weave " (merging) is the process of combining the odd and even fields of an interlaced video image to generate the entire non-interlaced frame for the progressive scan monitor.
Each method is simple to implement and has been around in the video industry for some time now, but each introduces some visual artifacts which leads to poor image quality.
" Bob " works well for images that are in motion but not for still images. If, for example, interlaced video is 480i (NTSC format), then the Bob method would just use the 240 scan lines of either the odd or even field and interpolate between the lines to generate a non-interlaced 480 line frame for the progressive scan monitor. The result is that the vertical resolution is decreased by half. Although this method works well, it can produce shimmering artifacts which create fuzziness and loss in detail in the vertical picture resolution. Artifacts are most obvious when text and stationary objects/logos are overlaid on live video. On some images objects may appear to bounce up and down. These artifacts may be worsened with frame rate conversion, such as converting from 60 Hz to 72 Hz.
" Weave " is perfectly adapted for still images because it uses all of the information available and thus generates the best possible image. While it works well with still images, it creates annoying feathering (combing) artifacts with motion video. Fast video and text scrolling can become unbearable. Commonly known as “feathering,” the vertical edges of moving objects in the image look like feathers. When an image is moving, any attempts to weave the fields will result in significant motion artifacts. Because the recording is performed in an interlaced manner, the two source fields that make up a complete frame are not recorded at the same time. Each frame is recorded as an odd field from one point in time, and then as an even field recorded 1/60 of a second later (for NTSC). A car moving at 60 mph will move by almost 1.5 ft. in this time, so that vertical edges in the odd and even fields will be separated by an equivalent distance. Simply combining fields causes the errors in the image called “combing” or “feathering” artifacts.
A simple solution to the problem of deciding whether to implement either the “bob” or the “weave” process is to switch between the two processes, depending on whether or not there is motion between the two fields. On a field-by-field basis, using a simple motion calculation, this technique would detect when the image is moving and implement the “bob” process and detect when the image is still and switch to the “weave” process. This sounds like a good solution to the “bob” or “weave” options that result in visual artifacts; unfortunately, most frames contain a mixture of both motion and still images, and whichever process is selected, artifacts are possible in those portions of the image for which the process is not optimized.
First introduced with DeltaChrome processors and further refined for Chrome S20 Series processors, S3 Graphics' Chromotion provides a more precise technology to determine the optimal scanning process, and provides superior de-interlacing.
Because most images contain a mix of still and moving regions, Chromotion's per-pixel adaptive de-interlacing technology looks at fields on a pixel-by-pixel basis and optimally determines whether to “bob” or “weave” on a pixel-by-pixel basis, based on the detected motion for that precise pixel. This advanced algorithm looks at pixels from the previous field, from the next field, and from the current field, and mathematically determines if there is motion. Thus it generates an appropriate, corresponding pixel for the current frame. Because this superior algorithm looks at the image on a pixel-by-pixel basis rather than on a field-by-field basis, it eliminates the visual artifacts in moving objects and preserves full resolution of non-moving portions of the screen. The result is a high quality crisp image, even on a HDTV progressive display.
| 
