• Blogs
• Tradeshows
• Partners
• White Papers
• Events

Edit Review — Algolith AlgoSuite

Feb 10, 2006 12:00 PM, Reviewer: Jan Ozer

A plethora of plug-ins for enhancing video footage.

I have a confession to make — not all of my video shoots go off perfectly. I know, I know, all of you out there get it right first time, every time, but unfortunately that's just not me. Matter of fact, when Ben Hogan said he only hit about four perfect shots per round of 72, he pretty much described my average when it comes to shooting interviews, concerts, and other events. Whether it's audio problems, inadequate lighting, funky white balancing, or a slightly shaky camera, something usually comes between me and audiovisual perfection.

Which is why I have an enduring affection for software that helps hide my screw-ups, whether it's Premiere Pro's shadow/highlight filter (backlight issues), Sound Forge's noise reduction filter (hum and background noise), color correction from a number of vendors, and the excellent Steadymove plug-in from 2d3 for After Effects (image stabilization).

Well, the newest arrow in my quiver of face-saving techniques is Algolith AlgoSuite ($1,185). This software includes modules for scaling, format conversion, anti-aliasing, and noise reduction, all presented as plug-ins for After Effects and Digital Fusion, with some modules available for Shake. Note that Algolith sells the format conversion and noise reduction plug-ins separately ($595 and $395, respectively), and offers one-week licenses for all tools for $59.

I tested on After Effects. The filter presentation proved complex, and rendering cycles were glacial. Still, if you're converting footage from SD to HD, this tool is an absolute must-buy. If your videos occasionally suffer from sub-optimal framing and low light induced chroma noise, AlgoSuite can also prove invaluable.

Overview

The complete suite incorporates nine separate plug-ins. Some you apply individually, while others are available as components of higher-level filters. For example, when you convert from DV to HD, you select the Format Conversion plug-in, which includes automatic and manual controls for noise reduction, de-interlacing, and scaling. Rather than describe each plug-in on a standalone basis, I'll identify the problem I was trying to solve with the suite, and then discuss which plug-ins I used in the process.

Let me start by saying that the downloadable manual, while not perfect, is absolutely essential. If you simply install the software and start working, you'll be in for a frustrating time. That's because some plug-ins work only on progressive footage, so you may have to convert from interlaced to progressive source first, then apply the plug-in, then convert back. In After Effects, Algolith plug-ins must be used at the beginning of the effect tree and shouldn't be “nested.” Reading through the application-specific sections of the manual will save you time and frustration.

Be prepared for some confusing error messages. When attempting to scale 4:3 DV footage (interlaced, 0.9 pixel aspect ratio) to 640×480 (progressive, square pixel) for encoding into Windows Media format, error messages advised me to interpret the footage as progressive and that the pixel aspect ratio had to be set to square, though intuitively these settings weren't correct.

Ultimately, I got all the filters working properly. But it was not without some head-scratching, control-wiggling, and multiple attempts. Given the lengthy rendering times for most processes, you should experiment with very short clips (15-30 seconds) and test the results for quality and workflow compatibility before pulling the trigger on your 60-minute video.

With this as background, let's dive into the projects.

Test 1: SD to HD conversion

Here I converted a 02:17 DV file with a 4:3 aspect ratio to 1280×720p using the Format Converter plug-in. This plug-in can convert between SD and HD or NTSC and PAL. Once I applied the effect, Algolith spawned three error messages dealing with progressive vs. interlaced, aspect ratio, and frame rate, which I sorted out by reading the manual.

When producing my test clip, I declined noise reduction and used the default de-interlacing option (Figure 1). Since my 4:3 source video footage didn't precisely scale to 16:9 1280×720p, AlgoSuite provides three configurable options: Variable, which stretches the image to fit the expanded frame, distorting the aspect ratio; Crop, which scales the image to match the horizontal resolution and crops the top and bottom of the frame; and Pad, which displays the entire 4:3 video with black bars on both sides. I chose Crop for my tests and rendered the HD video, which took nearly five hours (04:43.33) on my dual-3.06GHz Xeon computer. Then I removed the filter, zoomed to approximately 2X using After Effects' own scaling controls, which took a mere 15:38, and compared the results.

In this case, the AlgoSuite footage was clearly worth the wait. My test clip contained a number of discrete scenes with real-world and test footage-like resolution and color charts. In almost all clips there was a noticeable difference between the aliased footage from After Effects and the nearly pristine output produced by AlgoSuite.

This was particularly evident in footage containing many diagonal lines, such as the chip-fabrication machine shown in Figure 2. If you compare hard edges on the left with the equivalent section on the right, you'll notice the almost total lack of jagged edges, courtesy of AlgoSuite's Motion Adaptive Anti-Aliasing De-Interlacer (MAADI).

Even with a more natural image, like the ballerina shown in Figure 3, the difference in quality is quite striking, particularly in the hand and arm, but also in the hair and lips. Though you'll have to plan ahead to work around AlgoSuite's rendering time, the quality is almost certainly worth it.

Test 2: Zooming SD footage

The next test involved real world footage from an interview with a local congressman, Representative Rick Boucher. During the interview I framed the congressman far too loosely, particularly for video potentially bound for streaming delivery. I was editing the video in Premiere Pro. So first I tried adjusting the image there, zooming in to about 2X resolution and reframing the video using Premiere Pro's motion controls to maintain rule-of-thirds positioning.

You'll see in Figure 4 that the results were too fuzzy to be acceptable, though the process took a speedy 27 seconds for the 37-second sample. I duplicated the procedure in After Effects, which was slightly sharper, rendering in 52 seconds.

Then I applied the AlgoSuite Content Adaptive Scaling (CAS) filter, zooming the video to 2X resolution and accepting all defaults. I also applied the Anti-Aliasing Processor (AAP) and Multiple Type Noise Reduction (MTNR) filters, again accepting default settings. AlgoSuite took 37:50 to render, but produced the sharpest image by far, along with noticeably less fading than produced by After Effects or Premiere Pro. I may apply a slight blur filter to soften the image before production, but there's no question that AlgoSuite provided an extraordinarily useful corrective capability.

Test 3: Noise removal

My next tests involved several sequences (totaling a 42 seconds) that exhibited noticeable background noise, typically induced by sub-optimal lighting, a constant issue when shooting in the real world. Here I applied the MTNR filter, using default settings, which AlgoSuite rendered in a long 04:34:24.

Playing the processed and original files side-by-side revealed significantly less noise in the processed file, with no blurring or other artifacts that are often produced by noise reduction filters. To test whether the reduced noise would improve post-compression quality, I rendered both the original and processed files into Windows Media format at 320×240, 15fps, at 100kbps; into MPEG-2 format at 720×480, 30fps, at 8Mbps; and into Flash format at 640×480×30fps at 300kbps.

There was very little difference in the Windows Media files, perhaps because many codecs apply a noise reduction filter before compression or use similar techniques during encoding to discern between real motion and noise. That said, while the individual frame quality in the compressed files was similar, the file processed with AlgoSuite exhibited much less background noise in both clips, a subtle but noticeable improvement.

Of course, the value of noise reduction varies from clip to clip. Still, if you're expecting AlgoSuite's noise reduction filter to dramatically reduce visible artifacts in your compressed footage, you might be disappointed.

Test 4: Scaling for streaming

The final test involved scaling the original DV test file from 720×480 to 640×480 and to 320×240, a process typically performed when encoding for streaming. For this trial, I used the Motion Adaptive Anti-Aliasing and Content Adaptive Scaling for AlgoSuite, and, for the control video, After Effects' native scaling algorithms for the control video.

In the 640×480 trials, AlgoSuite's advantages were very evident, as shown in Figure 5, where the horizontal lines on the golf shirt are much sharper in the AlgoSuite clip. My test clip here was 02:47, which took 04:08:12 to produce in AlgoSuite, but only 06:42 using After Effects filters. The AlgoSuite video was also superior in 320×240 trials, though the disparity wasn't as great, probably because the process of scaling to half-height resolution naturally eliminates many interlacing issues.

Poor scaling and anti-aliasing hinders the performance of many standalone encoders, particularly Microsoft's Windows Media Encoder and Macromedia's Flash 8 Video Encoder. If you're rendering your streaming files in one of these tools, scaling and deinterlacing in After Effects should significantly improve your results, particularly with the AlgoSuite plug-ins.

Summary

To recount, AlgoSuite's scaling tools produced the most advantage when scaling to a higher resolution, either within a DV file (fixing the Boucher video) or when scaling to a higher resolution, as in SD to HD conversions. Performance was also quite impressive when scaling from DV to 640×480, but less so to 320×240 resolution.

AlgoSuite's noise reduction feature dramatically reduced background noise without distortion or other artifacts. Though this didn't reduce the artifacts in my highly compressed streaming samples, the reduction in noise in higher bit rate samples, including MPEG-2 and Flash, visibly improved post-compression appearance.

Between corrective capabilities like high-quality re-framing and noise removal, and production features like scaling DV to HD or to lower resolutions for streaming, AlgoSuite has a boatload of goods that most production houses can use. A free trial would be nice, but Algolith's weekly license provides an economical way to determine if the filters work well in your environment.

---

bottomline

Company: Algolith
Montreal; (514) 335-9867
www.algolith.com

Product: AlgoSuite

Assets: Beats After Effects filters on image quality in a variety of tests.

Caveats: Render times run into the hours for minutes of footage.

Demographic: Production houses that need fine control over noise reduction, anti-aliasing, and image scaling.

PRICE: $1,185

© 2008 Penton Media, Inc.