Axis Zipstream Claims Average 50% Savings

By Ethan Ace, Published Mar 25, 2015, 12:00am EDT

Axis claims that its new Zipstream compression technology can "lower bandwidth and storage requirements by an average 50% or more."

That is quite an audacious claim for such a fundamental element of video surveillance systems.

In this note, we explain how Zipstream works and how it matches up against other bandwidth 'saving' technologies and rival products:

  • Dynamic Scene Compression - How It Works
  • Dynamic I Frame Interval - What It Does
  • Axis Best and Worst Case Zipstream Scenarios
  • Axis Availability Restrictions
  • Zipstream vs H.265
  • Zipstream vs Vivotek Smart Stream and Arecont Bandwidth Savings Mode
  • Zipstream vs Ambarella
  • Zipstream vs Dahua and Hikvision
  • Zipstream Marketing Impact

Background: Axis Zipstream press release and Axis Zipstream whitepaper [link no longer available]

Bandwidth Variations Huge

Bandwidth variations, even across cameras from the same manufacturer with the same CODEC, resolution, frame rate and compression can be huge. This is important to appreciate and understand how meaningful or not Zipstream will be.

The image below shows two Axis cameras, the Q1604 and M3004, both set to 720p, 10 FPS, quantization of ~28, across a 20' horizontal FOV. Despite these factors being standardized, the Q1604's bitrate was 488 Kb/s while the M3004 was 1.32 Mb/s, nearly 3x the bandwidth.

Additionally, factors such as sharpness, contrast, WDR performance, etc., reveal differing levels of details in different scenes, as we show in our Advanced Camera Bandwidth Test Results. Due to these variations, predicting how much savings Zipstream will provide in all these situations is near impossible.

Dynamic Scene Compression - How It Works

Join IPVM Newsletter?

IPVM is the #1 authority in video surveillance news, in-depth tests, and training courses. Get emails, once a day, Monday to Friday.

All H.264 is compressed, and the amount of compression is controlled by the compression / quantization level. In Axis cameras, this is a range from 0 (least compression) to 100 (most compression). This maps to H.264 quantization levels of 0 to 51 (for background, see our Video Quality / Compression Tutorial).

Normally, a single compression level is applied to an entire video stream. For example, a default Axis camera has a compression level of 30. The whole scene - whether it is a white wall or a person's face has the same compression level.

The first of the two key Zipstream features, Dynamic ROI adjusts compression / quantization levels based upon scene content. Moving objects such as people or vehicles may have a quantization level of 27 applied, while the background of the scene such as walls and pavement are more compressed, e.g., a quantization of 35. 

The benefit of using different compression levels for different parts of a stream is that 'bits' are 'wasted' on certain portions. For example, you generally want low compression on a face but high compression on a blank white wall is typically fine and saves bandwidth / storage.

The key issue with this technique is how accurate the technology is in deciding what parts of the scene can be applied a high compression without losing meaningful details.

Dynamic I Frame Interval - What It Does

In addition to adjusting compression based on scene activity, Zipstream also adjusts I frame interval. For example, in a still scene with no motion, an I frame may only be sent every 5 seconds or more, since no major changes have occurred in the scene. When someone enters the scene and walks through, however, the camera adjusts the interval to a shorter duration, sending an I frame or more each second. 

Axis informs us they are including a "max GOP length configuration that limits how long the dynamic GOP can be." This provides a brake to ensure that very long I frame intervals do not cause problems for VMS systems that depend on regular I frame intervals for search and playback.

Axis Best and Worst Case Zipstream Scenarios

In Axis' Zipstream whitepaper, here are their claims for benefits across different scenes:

Our interpretation is that:

  • Narrower FoVs, like the retail scene above, are hard to compress much more without material quality loss, because they contain lots of details both of objects and people walking through.
  • Wider FoVs, like the parking lot above, are easier to compress more without significant quality loss, because the objects are relatively small / undetailed.
  • Night time recording is the most likely to waste bandwidth due to compressing noise artifacts, and therefore, offers the highest savings potential. Related, see: Testing Bandwidth vs Low Light and Tested: Lowering Bandwidth at Night is Good.

Axis Availability Restrictions

Almost none of Axis' existing cameras will support Zipstream, severely restraining its usability in the short term.

Zipstream is limited to Axis cameras running ARTPEC-5, which is Axis' newest encoding chip. The only existing cameras that can be upgraded to Zipstream are AXIS Q1615, AXIS Q1635 and the AXIS Q35 Series. However, Axis is releasing a slew of new in April and, of course, regularly releases new models.

Firmware to upgrade existing cameras that are Zipstream upgradeable is scheduled for release on April 9, 2015. We plan to test the Q1615 when this is done. 

Zipstream vs H.265

We see high demand for H.265 but do not expect Axis to adopt it this year. To that end, Zipstream could act as counterforce against smaller rivals rolling out H.265, which is already happening and will surely accelerate throughout 2015.

Note that H.265 key marketing claim is 'up to 50 savings', very similar to Axis claim with H.264. To that end, we anticipate the Axis argument being, 'Why switch to a new codec which lacks VMS support and introduces risks of higher decoding CPU use, when you can stay with the brand you know/trust and a proven codec - H.264.'

Zipstream vs Vivotek Smart Stream and Arecont Bandwidth Savings Mode

Zipstream shares some similarities with Vivotek's Smart Stream, as both dynamically adjust compression on objects in the scene to reduce bandwidth. In our test of Smart Stream, reductions of 15-30% in well lit scenes and up to 50% in low light, both short of Axis' 25% and 90% predictions in these scenes. Additionally, in the Vivotek camera tested, average bandwidth consumption without Smart Stream was far higher than average, so while these reductions were actual, bandwidth remained fairly high.

Arecont's Bandwidth Savings Mode, another bitrate reduction feature, does not adjust compression or GOP as Zipstream does, but instead removes some details from the scene, adjusting greyscale and reducing the number of colors displayed. We found bandwidth savings from 12% to 70% depending on the scene, but multiple image quality issues due to this mode. 

Zipstream vs Ambarella

Ambarella is the biggest encoder chip provider for high-end IP cameras and their chips (like the S2 and the upcoming S3 H.265) largely shape bandwidth consumption / savings for cameras using them. Ambarella’s chips support optimizations they call “SmartAVC” that extend usable bitrate range downwards, that they claim to be as low as 300 - 750 Kbps for Full HD video. Ambarella customers, of course, have different names for these features.

Two Ambarella customers, Bosch and Samsung, did very well in our 2014 IP Camera Bandwidth / Storage Shootout. We plan to test Axis Zipstream against such cameras.

Zipstream vs Dahua and Hikvision

Neither Dahua or Hikvision market any special bandwidth savings. To that end, Axis may be able to position Zipstream as a reason to pay a premium for their cameras, justifying the price gap / increase of Axis relative to their biggest Chinese competitors.

Zipstream Marketing Impact

Far more so than any surveillance manufacturer, given their size and partner numbers, Axis has the greatest ability to move the market with their initiatives (e.g., ONVIF and LightFinder). To that end, we think that Zipstream could spur rivals to launch their own compression optimizations / marketing buzzwords. And in the interim, as LightFinder did for the first 18 months after release, could provide them a meaningful competitive advantage.

Ultimately, though, a significant portion of Zipstream's impact will be based on how much it can really decrease bandwidth without sacrificing storage. We expect to text the Axis Q1615 when new firmware is released in April 2015.

3 reports cite this report:

New Products Spring 2015 on Apr 12, 2015
After very poor new releases in 2014 (see Spring 2014 and Fall 2014...
Axis ACC v3 + SanDisk OEM on Apr 09, 2015
The most talked about new release in 2012 was Axis Camera Companion, aiming...
Hikvision Launching Smart H.264 and H.265 on Apr 08, 2015
The CODEC wars are heating up. Axis made waves just weeks ago with their...
Comments (22) : Subscribers only. Login. or Join.
Loading Related Reports