Multimedia
Blogs
Forum
Affordable HD
Whitepapers
Newsletters
Events
Advertisers
DCP Directory
MillimeterWorking with Legacy Video Formats: Part 2
Jun 26, 2006 12:22 PM
In Part 1 we looked at how we can incorporate legacy SD video into HD productions. We closed Part 1 by examining how comb filter-based Y/C separators work — and how they do not meet our needs.
A two-dimensional, three-line, adaptive comb filter provides improved comb filter performance. It compares each pixel to the immediately adjacent pixels on the current line as well as adjacent pixels in the upper and lower lines. Based on these pixels' hue and saturation, digital logic causes the Y/C separator to, on a pixel-by-pixel basis, switch between operating as a pure notch filter (for minimal color artifacts) and operating as a pure comb filter (for maximum picture resolution).
While the upper and lower lines from the same field are close temporally, they are not spatially close. The spatially closest lines are in the previous field. A 3D, three-line Y/C separator employs field memory. Because its chroma logic works with pixels that are spatially close, it provides better picture detail with fewer chroma artifacts.
While comparing the current line with the upper and lower line from the previous field usually provides optimum separator performance, when an object is moving rapidly, it does not. By incorporating motion-sensing logic, a 3D Y/C separator can automatically use the previous field or current field.
Once chroma has been separated from luma, each sub-signal must be cleaned of noise. Picture noise is uncorrelated information. Digital Noise Reduction (DNR) delays video one line period and adds it to the current line. Luminance noise is canceled and the signal augmented. A 2D, two-line, DNR circuit causes a slight blurring of vertical detail. For this reason, DNR is available with the same advanced modes as are Y/C separators: 2D three-line, 3D three-line, and motion-sensing 3D three-line.
There are two types of chroma noise — AM (Amplitude Modulated) noise where each pixel has random saturation and PM (Phase Modulated) noise where pixels have a random, varying hue. The same types of digital noise reduction circuits that reduce luminance noise can be applied to a chroma noise.
There is one other nasty you may need to deal with: “head-switching” noise. Found in the lowest six to eight lines of video from an analog VTR, this noise can be eliminated by simply masking the top and bottom eight lines. Alternately, you can crop the video, and then slightly scale it. However, because this increases the amount of scaling required, picture quality can be decreased.
Some VHS and S-VHS tapes will carry a HiFi audio track that offers high-quality stereo sound. However, if it's played on a VTR with worn HiFi audio heads, you may hear a 15kHz, head-switching noise. Hopefully, your NLE has a 15kHz notch filter so you can reduce the whine.
In my experience, the VTRs that come the closest to meeting most of our requirements are the Panasonic AG-7700/AG-7600 and AG-7650/7750, plus the JVC BR-S822U/BR-S622U (VHS/S-VHS) decks. If possible, use a 22-series VCR that has both an SA-T22 TBC and an SA-N22 DNR option. Other decks include the Sony BVU-950 (U-matic) with a TBC connected to the DUB port, a Sony SVO-5800 (VHS/S-VHS), and a Sony EVO-9850 (Video8/Hi8). At a much lower price point, the JVC BR-S378U and Panasonic AG-DS555 VHS/S-VHS decks work well.
Two more points. If you must choose between a path that adds noise or one that adds cross-luma or cross-color artifacts, choose to add noise. Noise can be removed — artifacts cannot. For the same reason, unless the deck you use has component analog output, you must use an S-Video connection.
In terms of a software solution, Compression Master V4 (from Popwire) has four filters that support frame-rate conversion, resizing, de-interlacing, and noise reduction. The program runs under OS X and comes as a Universal application. I have been using it for a month and recommend it highly.
When working with film-based media, Compression Master’s frame-rate filter offers both “Telecine (23.98 => 29.97)” and “Inv. Telecine (29.97 => 23.98)” functions.
To scale video up or down, Compression Master’s resize filter offers several scaling options. When using the resize filter, you have several options: Automatic (bilinear when downsizing and Bicubic for upsizing), Bilinear, and Bicubic, and Nearest Neighbor.
Compression Master has a powerful De-interlace filter that provides a large selection of operating modes. These include: Complete De-interlace (de-interlace whole frame), De-interlace Interlaced Frames (frames in the material that are interlaced are deinterlaced), De-interlace Moving Areas (de-interlace the moving parts of each video frame).
When employing legacy video in HD productions, removing noise is a critical need. Compression Master provides a Noise Reduction filter that supports three techniques: Median, Average, and Temporal.
Another conversion solution comes from Digital Anarchy. ReSizer 2 is a plug-in that intelligently scales video. The plug-in works in Adobe After Effects and Premiere Pro, as well as Final Cut Pro. ReSizer 2.0 has three scaling algorithms. Users also have control over how much these algorithms sharpen or smooth the footage. For working with SD interlaced footage, ReSizer 2.0 has a Deinterlacer that uses motion compensation. Multiple frames of interlaced footage are analyzed, and the missing scan lines are intelligently reconstructed.
By using an advanced VTR plus powerful software, you should be able to maximize the quality of legacy SD video used within HD productions.
