De-interlacing explained

What’s de-interlacing?

De-interlacing is the process in which interlaced video is converted into a progressive scan format. We really need to dedicate a whole article for this, but the general principle is that de-interlacing is the critical determinant of final video picture quality when dealing with 1080i content. The difference between a $2000 DLP front projector and $15,000 one is largely the video processor that does the de-interlacing.

You probably already know that 1080i HDTV has 1920x1080 pixels of resolution. Since the video is interlaced, of those 1080 rows only 540 lines are recorded at each step in time: first the odd rows and then even rows. The basic problem is that in interlaced video, the two fields are recorded a fraction of a second apart in time. This means that if you simply try to display the two fields on a monitor, you will see an artifact known as feathering.

Because the moving object (the blue damselfish) had moved a significant distance in that 1/60th of a second. On the other hand, the coral in the background were not moving and so they remain sharp. Since feathering artifacts are distracting, the challenge is to finding a way to remove these artifacts. The goal of de-interlacing is to somehow update the entire screen every 1/60th of a second even though only half of the screen is being updated. That is, you’re NOT going from 1920x1080i at 60Hz to 1920x1080p at 30Hz, you are trying to go from 1920x1080i at 60Hz to 1920x1080p at 60Hz. You’re generating twice the data, hence the phrase “line-doubling.” Doing this is no easy feat, and companies like Faroudja have won technical Emmy Awards for their algorithms.



The easiest approach to solving this problem is to ignore half of the fields. Although the video is updated at 60 Hz, only half of the screen (one set of fields) is updated each time. Through “conventional bob” de-interlacing, you alternate between displaying only the odd fields and then only the even fields. Unfortunately, this means that you’re losing all of the detail and resolution in the static elements of the screen. There’s no reason why you couldn’t have shown all of the fields from the coral in the background…

So, with interlaced video, anytime there’s motion in a scene, you’ll lose resolution. That’s just the nature of it. De-interlacers that apply the “conventional bob” algorithm end up losing as much as half of the detail and resolution from the original source. That 720x480i DVD becomes sequential frames of 720x240 and that 1920x1080i HDTV becomes sequential frames of 1920x540p. So, when it comes to different deinterlacing techniques, it all boils down to figuring out how to minimize the amount of pixels you lose.

The HD Challenge

Most manufacturers have figured out clever ways to de-interlace standard definition DVD-quality video. The problem arises with 1080i HDTV content such as what you will see on HDNet, CBS, NBC, PBS, UPN, HBO, Showtime, INHD, Discovery Channel HD, and the WB (Fox, ABC, and ESPN use 720p). Since 1920x1080 represents six times the data, most deinterlacers, including those from Faroudja and the PureVideo in the GeForce 6600 are unable to intelligently deal with 1080i sources. For the majority of de-interlacers, 1080i HDTV is de-interlaced by ignoring one of the fields during each stage of processing. This means that 1920x1080i image is converted to 1920x540p before it goes to 1920x1080p. You can lose half your spatial resolution when considered on a frame by frame basis. This is an industry-wide problem. That ultra-high-end DLP for $5000? When you feed it a 1080i source, it’s likely using one only one-field with 540 vertical lines to convert 1920x1080i into 1920x540 before going to1280x720p. Even most high-end 45” 1920x1080 LCD flat panels will take 1920x1080i down to 1920x540p before going up to 1920x1080p.