Progressive Scan Technology
Dear Gary: I have noticed that virtually every progressive scan DVD player possesses a different progressive scanning technology. Sony’s is different from Toshiba’s; Pioneer’s is different from Panasonic’s. Why must this myriad of methods be? Why can’t all standalone progressive scan DVD players follow the example of computer software DVD players (e.g., PowerDVD) and weave? It seems that all a progressive scan player needs to do with film-based material is to weave the two interlaced fields that come from the same frame (same moment in time) and then perform 2:3 pulldown. In my opinion, progressive scanning seems so simple. So why does it have to be so complicated?
Jason Steg, Chapel Hill, North Carolina
Video Technical Editor Greg Rogers
Conceptually, it is fairly straightforward to convert interlaced video from film sources to progressive scan video, a process known as deinterlacing. But there are a number of practical problems that complicate the process and may create imperfect results. Progressive scan DVD players, and standalone line doublers or scalers, use a variety of off-the-shelf and custom deinterlacing technologies to balance cost and optimize performance. The telecine process used to convert film to video separates each film frame into two interlaced fields, and then one redundant video field is repeated for every other film frame. This creates a sequence of three video fields followed by two video fields for every two film frames, a process known as 3:2 pulldown. The result is 60 field-per-second interlaced video from 24 frame-per-second film. Ideally the 3:2 pulldown sequence would repeat uninterrupted throughout a movie. However, editing during the telecine process, or subsequent video processing, can break the 3:2 cadence and create problems during deinterlacing. The essential and complex element of deinterlacing film-source video is to identify and lock on to the 3:2 pulldown cadence. Once the cadence is recognized it is straightforward to remove the redundant video fields and merge (called weave by computer people) the appropriate video fields to create progressive video frames that mirror the original film frames. This is known as the inverse-telecine process. It creates progressive frames free of deinterlacing artifacts because it merges back together the content of each film frame that was separated into fields in the telecine process. Since this would result in 24 frame-per-second progressive video, the final step is to repeat the reassembled progressive frames in another 3:2 sequence to generate 60 frame-per-second progressive video. When video is MPEG-encoded for a DVD, the video data between pairs of fields is compared to identify the redundant fields and the 3:2 pulldown cadence. The redundant fields are removed and flags are asserted in the MPEG bit stream that tell the DVD player’s MPEG decoder to repeat the necessary fields to recreate the 3:2 pulldown sequence. This reduces the amount of data that must stored on the DVD. If the MPEG encoder fails to detect a redundant field because of noisy video or a broken cadence from editing, it will encode the field rather than asserting the appropriate flags. This has no effect on the interlaced output of a DVD player other than using a few more bits on the disc. But it may create serious deinterlacing problems for a progressive scan DVD player that relied only on those flags. The first generation of standalone progressive scan DVD players used deinterlacing chips designed for line doublers, following the digital interlaced output of the MPEG decoder. They had no access to the MPEG bit stream flags and instead compared the data content between fields to identify and lock on to the 3:2 cadence. Their ability to provide artifact-free and glitch-free pictures from movies varied from product to product based on the strategies they employed to detect and handle cadence breaks from editing. The latest standalone progressive scan DVD players and computer software DVD players, read the MPEG bit stream flags to identify the 3:2 cadence. There are actually several flags (repeat_first_field, top_field_first, progressive_frame) involved in this process. Unfortunately, there are various types of flag errors from the MPEG encoding process that would lead to merging the wrong fields. So the latest players, like the Sony DVP-S9000ES, use a combination of reading flags and examining the video data to lock on to the cadence more quickly, and to minimize visible artifacts when flag errors occur or the 3:2 cadence is interrupted. The strategies employed to handle cadence interruptions and flag errors distinguish one DVD player or standalone processor from another. Even though they have no access to the MPEG bit stream, some external processors, such as the Faroudja scalers, show fewer visible artifacts on problem material than the internal deinterlacing in many DVD players. This shows that the diverse strategies for handling these problems are reflected in the performance of the products. Finally, deinterlacing of original video source material is an entirely different game. There are no ideal algorithms that can provide perfect results. There is no 3:2 pulldown in use, so inverse-telecine processing does not apply. Quality is often a function of implementing more complex algorithms such as adaptive directional interpolation. In this case, standalone video processors and PCs with custom digital video processing ICs, or high-speed DSP chips, will have an edge over consumer DVD players.
You can E-mail Widescreen Review @ mailto:editorgary@widescreenreview.com