• Blogs
• Forums
• Tradeshows
• Partners
• White Papers
• Podcast Show Notes
• Events

 

Just what is 1080i? Part 4

Mar 27, 2006 6:49 PM

Alert: 720p editors need to read Part 4, as do those shooting 1080i interlaced video that will be transferred to film. The moment legacy interlaced video is dropped into a 720p timeline, the editor will have to decide how it should be converted to progressive. Likewise, when editors complete a 1080i interlaced production, they face the issue of how they will convert interlaced to progressive — whether they do it or the film lab does it.

In the last Newsletter, I promised to explain why 48 percent of current HDTVs cannot display more than 540 lines of 1080i video. This startling fact was revealed by Gary Merson’s tests on 54 current 2005 model “HD resolution” HDTVs of all types (Home Theater, March 2006). These HDTVs were progressive displays with 720 rows, 768 rows, or 1080 rows.

Merson used a moving 1080i SMPTE 133 test pattern. If 1080 lines of vertical resolution were seen, the HDTV passed the test. If not, the HDTV set failed.

If you have an earlier model HDTV, the odds are very high that it would fail the test, whereas if you have a 2006 model the probability is higher that would pass. (Obviously, 720p video need not be de-interlaced.)

You might assume that manufacturers who promote the virtues of high-resolution of HDTVs — especially those with 1920x1080 panels — would build sets whose circuitry supplied video with adequate vertical resolution to fill these panels. Unfortunately, it is common practice for manufactures to promote certain specifications — 1080i and 1080p come to mind — while building these products with components that do not support these specifications.

The inferior part inside the failing HDTVs is the de-interlacer — a critical circuit that does the very hard job of converting interlaced video to progressive video. There are multiple ways to accomplish de-interlacing, these methods fall into two categories: non-adaptive (bob, weave, and 2D FIR) and adaptive (frame-adaptive, region-adaptive, and vector-adaptive).

Why is de-interlacing using an interpolator called bob? To some, the constructed image seems to vibrate up and down, hence the term “bob.” An interpolator generates each target line from pixels in the lines directly above and below target lines. Diagonals are not handled well because an interpolator cannot create pixels in the lines that would yield smooth diagonal lines since there are no pixels above and below the locations (yellow cells) where they should be placed.

As shown in the Diagram above, the generated pixels (red dots) create a frame with no interlaced artifacts, but with many jaggies. A second-stage, anti-alias filter is necessary to reduce jaggies. Effective vertical resolution is that of a single field — 540 TVL.

Weave uses both even and odd lines, which are woven into a single 1080-line frame. Because information from different moments in time may be combined, moving objects will have combing on their edges. A second-stage, isotropic filter is critical to blending pixels at the edges of moving objects in order to reduce combing. Static resolution tests will indicate an effective vertical resolution of 1080 TVL.

After one field is discarded, the remaining 540-line field is vertically scaled to a 1080-line frame using a 2D FIR filter. A 2D FIR filter can have a small or large number “taps” where each tap is a sample (blue cells). Current filters have up to 1024 taps, which would support a 32x32 window around each target pixel (yellow cells).

The generated pixels (red dots) yield a frame with no interlaced artifacts and relatively few jaggies. Vertical resolution will be approximately 750 TVL because 2D FIR filters increase resolution approximately 1.4X.

Frame-adaptive de-interlacing uses logic to measure motion between fields. For static frames — frames with little or no motion — weave is employed. For dynamic frames — frames with motion — either a 2D FIR filter or a bob interpolator is used to generate a new frame.

Region-adaptive de-interlacing uses logic to measure inter-field motion within regions. The smaller the region, the more total image resolution is maximized. Weave is used for static regions. For dynamic regions, either a 2D FIR filter or a bob interpolator is used to generate a new frame.

A vector-adaptive interpolator uses memory to hold four fields. Logic measures motion between fields in memory. For static video, weave is employed, and vertical resolution will be 1080 TVL. For dynamic video, samples come from the current plus a previous and/or a future field. This creates a 3D vector interpolator that supports up to four vectors within a field, plus up to nine into the preceding and following field(s). Dynamic video will have approximately 960 TVL of vertical resolution.

The HDTVs that failed the test use a bob de-interlacer that limits vertical resolution to 540 lines. So even with a 1920x1080 panel, total resolution is only 1920x540—1,036,800 pixels. You might be surprised that 720p video will fill the same panel with 921,600 pixels. Thus, there is about a 50/50 chance the 1080i video you swear looks so much sharper than 720p video on your HDTV, has only an insignificant, 10-percent greater resolution.

In case you see these test results as confirming the superiority of CRT-based devices for viewing 1080i video, think again. In the next, and last, Newsletter coverage of 1080i, we will look at CRT-based HD displays.

Continue the discussion on “Crosstalk” the Millimeter Forum.
© 2010 Penton Media, Inc.