IP Camera Manufacturer Compression Comparison

Author: Derek Ward, Published on Feb 21, 2014

Compression is very important. While resolution gets the attention, compression is critical and can be a silent killer - both for quality and bandwidth.

Regardless of resolution, all surveillance video is compressed. And even if 2 cameras have the same resolution, their compression levels can be much different. [See our compression / quality tutorial for background.]

Thankfully, compression in H.264 is standardized on a scale of 0 to 51, as shown in the image below:


However, camera manufacturers almost never disclose Q levels used. Instead, they use a variety of homemade scales and naming systems. Here is a sample of ones we tested inside:

So you can have 2 manufacturer's cameras with the same resolution but significantly different compression levels, and therefore varying image quality and bandwidth consumption.

An industry first, IPVM has analyzed each of these manufacturers and answered these key questions:

  • What is the real H.264 quantization level for each camera manufacturer's default settings? How do they vary? Who defaults the lowest and highest?
  • To normalize the H.264 quantization levels so that each manufacturer had the same compression, what camera settings should be used?
  • How does the range of compression levels used for each manufacturer map to H.264 quantization levels?
  • What is the impact of bandwidth as H.264 quantization / compression levels are varied for different manufacturers?

If you really care about image quality and optimizing bandwidth / storage use, this is a critical report.

*********** ** **** *********. ***** ********** **** *** *********, *********** is ******** *** *** ** * ****** ****** - **** for ******* *** *********.

********** ************, *** ************ ***** ** **********. *** **** ** * cameras **** *** **** **********, ***** *********** ****** *** ** much *********. [*** ************** / ******* *********** **********.]

**********, *********** ** *.*** ** ************ ** * ***** ** 0 ** **, ** ***** ** *** ***** *****:


*******, ****** ************* ****** ***** ******** * ****** ****. *******, they *** * ******* ** ******** ****** *** ****** *******. Here ** * ****** ** **** ** ****** ******:

** *** *** **** * ************'* ******* **** *** **** resolution *** ************* ********* *********** ******, *** ********* ******* ***** quality *** ********* ***********.

** ******** *****, **** *** ******** **** ** ***** ************* and ******** ***** *** *********:

  • **** ** *** **** *.*** ************ ***** *** **** ****** manufacturer's ******* ********? *** ** **** ****? *** ******** *** lowest *** *******?
  • ** ********* *** *.*** ************ ****** ** **** **** ************ had *** **** ***********, **** ****** ******** ****** ** ****?
  • *** **** *** ***** ** *********** ****** **** *** **** manufacturer *** ** *.*** ************ ******?
  • **** ** *** ****** ** ********* ** *.*** ************ / compression ****** *** ****** *** ********* *************?

** *** ****** **** ***** ***** ******* *** ********** ********* / ******* ***, **** ** * ******** ******.

[***************]

Test ********

***** ******* **** ******, *** ****** ** *** ******** *** * ********* ** **** ****.

Key ******

***** *** *** *** ****** ** **** ******:

  • ******* ************ ***** *** ** **** * ***** ** ** to **. *** ******* ***-**** ********* ** *** ****** ************ of *** ******* ****** ** * * ** **, *** the***********-*********-* ********** **** *** ******* ************ ** **. ***** ** chart ************ **** ******'* *********** ******* *************:

  • ********* ** ******** ******** ** ************ *****. ** * **** of *****, ***** **** * ** ** * ** *** resulted ** * **-** ********* ********, *** **** * * 34 ** * ** *** ******* **-***, ******* ** ************.
  • **** **** *** *** ******* (**** ** ******** **-****) ***** ***** ** ******** ****** *********** ********, **** *** camera ************* ******** *********** ******, **** **** *** **.
  • ****** ************* *** ********* ************ *** *********** ********, **** ****** which **** ***** ** *** ****** ********* ** ***** *************' compression ********. **** *** ***** ********* **** ****** ** *********** multiple ******* ** *** ************ *****.
  • ******** ****** *** ***** ********* **** *** ******* ******* ********* things. *** *******, ***** *** **** **** **** * ****** when *********** ** *** ** "*", ***** ********, ****, *****, and ******* **** *** * ****** **** *********** *** ****** to "*".

Camera *********** ******

**** ***** ***** *** *********** ************ ***** **** ****** ******* given *** *********** ********:

** **** ********** **** ******'* *********** ******** ** ******* * quantization ** ~**. **** *** ** ****** *** ***** ******* to ********* *********** ******** ** ***** *** *******, ****** ** a ***** *** ******* ****** **** *** **** ************.

Compression ****** ** *********

** ***** ** *** *** ********, ** **** ****** *** relative ******** ** ********* ** *********** ********* *** ***** ******* increases. * ** ** *** * ** ** ********* *** approximate ********* ******** **** * * ** **. **** ****** be ******** ** * **** ** ***** ****, ** ** acknowledge **** ********* ** ********* ****** *** ***** ****** ******** also ****** *********.

Setting *********** *** ********* *******

***** ** * *********** ** **** ******'* *********** ******** ****** their ********** *** **.

****

*** **** ***'* *********** ******** *** ** *** ** *** compression *** ***** "*****", ** **** ** ******************. **** **** * ******* ***** **** *** (**** ***********), medium, *** **** (***** ***********). *** **** *** ******** ** "medium" *******.

********

********'* *.**-**-** *********** ******** *** ** ******* ***** *** *********** and ***** **** *** ** ***** *** **, *** ** uses * ******* ****** ** "*" (****** ***********) ** "**" (highest ***********). *** ******** *.**-**-** ******** ******* ** "*".

****

****'************** *** ***** ***** *** ***** ****** ******** ***, *** users *** ****** * *********** ****** **** "*" (***** ***********) to "***" (**** ***********). *** **** ***** ******** *********** ** "30".

*****

***** ****** ***** ** ****** ********** *********** ** *-******, *-******, and ***** *********. *** ***** ***-**** ******** ***. *-***** ** to "****"

*****

*****'* *********** ******** *** ***** ** *** ***** ***, *** the "*******" *** ** *** ******* "*" (**** ***********) ** "6" (***** ***********). *** ***** *** ******* ******** ** ******* "4".

*********

*********'* *********** ******** *** ***** *** *****/***** ***, *** *** be *** ******* "******" (**** ***********) ** "*******" (***** ***********). The ********* **-*********-* ******** ***** ******* ** "******".

*******

*******'************** *** ***** ***** ***** ***** ******* ***, *** ***** from "**** (*)" (***** **********) ** "***** (**)" (**** **********). The ******* ***-**** ******** *********** ** "*".

****

** ****'* ******************, *********** ** *** ** *** ***** ******* * ******** menu, ***** ****** **** "*" (**** **********) ** "**" (**** compressed). *** **** ***-***** ******** ***** ******* ** "*".

Test ********

***** ******* **** ******, *** ****** ** *** ******** *** * ********* ** **** ****.

Comments (11)

Interesting.

Just mention that some manufacturers you forgot here, offer in addition to "pre defined quality values" some Manual settings where you can enter 0 to 51 customized values...so you get "settings for dummies " or "more accurate professionnal settings"

All this impact bandwidth IF you chose VBR output. If not, the output will be stable but you'd better chose the good CBR level to stand your Fps x resolutionxQuality choices. If not, the quality/Compression will become even worst (most of the time you keep the frame rate and downgrade quality ) This is whatoften happen by night ...and dark times ...brrrr

So you first think you will get a 28 quality , but finally you get a 35 or 40 and some pixelised fps where you can not identify your target... oups

Yes, this presumes the use of VBR, which is the typical default in IP cameras today. As you say, if you use CBR, by definition, bit rate will remain fixed but quantization level will vary depending on the scene.

Background: CBR vs VBR: Surveillance Streaming

Another great test, very informative.

John it would be very intersting to do a survey on who is using VBR or CBR. are you sure VBR will win in North America ? and worldwide ?

Most guys I know put CBR.. VBR +Cap is close to CBR but even worst if the limit is too low because hyper compressing before limitA part of Axis other vendors propose Fixed Quality, but most without a Cap, so it's seen as too risky for your storage, especially by Night.

Last survey with 60% VBR is strange .... next survey should ask also which brands are deployed...

I am pretty sure all 8 cameras in this test default to VBR (or VBR + a high cap).

You mean a survey of integrator selection (i.e. VBR vs CBR) or manufacturer defaults?

I think CBR is a really bad move for almost anyone and VBR + Cap is better. VBR + Cap allows the camera to stream at lower bit rates when the scene is less complex / hard to compress yet go up to whatever high point you choose. By contrast, CBR is wasting bandwidth for all times except for when the scene needs that set bit rate.

Derek, truly a commendable addition to the IPVM canon!

Thanks for the quiz, and here's a couple questions for you in return:

How was the Q measured, what tools, what method (mean average)?

Do you have a high confidence level in these numbers?

Mainly the section on Compression Impact on Bandwidth, arguably the most critical section, seems to present more questions than it answers.

I understand and acknowledge your disclaimer regarding the dangers of generalizing the data, but even granting that, there are some wildly divergent data points that should be explained if at all possible.

Take the Sony and the Samsung, witb the Samsung stream only getting slightly smaller with a q change of 6 where Sony gets almost 4x as small? These cameras are roughly equivalent hardware wise at least, so what is really going on? Like the frame rate changing? Or the profile?

More data would be helpful here to identify anomalies, but strangely half the cameras are missing from this chart without any explanation, and the ones we have left are divided between "Sony" behavior and "Samsung" behavior.

Any thoughts welcome.

Hello Rukmini Wilson,

Q was measured using AVInaptic, a free program that analyzes quantization, which we have been using for quite some time now.

As for your questions regarding the bandwidth chart, we wanted to show a trend as a rule of thumb that there is a strong likelyhood that decreasing compression or Q levels will increase bandwidth (Noting that compression is not the only factor in determining bit rate). No other settings such as FPS, resolution, or other codec or camera settings aside from compression/quality levels were changed via the camera's web UI.

Camera selection for the bandwidth section was random, and given the plethora of variables on the impact different cameras and manufacturers have with regards to bandwidth, 4 were chosen.

Could you please expand on what you mean between "Sony" behavior and "Samsung" behavior? What about "Dahua" and "Hikvision" behavior? Those 2 were included as well, no?

Hello.

Could you please provide a brief description of a procedure with which you obtained quantization level? You took 'average DRF'? What kind of files you put in AVInaptic?

Cheers.

See: How to Measure Video Quality / Compression Levels for full explanation of this.

Thanks John - everything clear now.

Thanks for the quick response Derek!

"Sony" behavior is: cameras which show dramatic (4X, 3X) reduction in bandwidth from q of 28 to 22

"Samsung" behavior is : cameras which show pathetic (1.2X, 1.6X) reduction in bandwidth from q of 28 to 22

The are two "Sony" types are the Hika and the Sony itself. The two "Samsung" types are the Dahua and the Samsung itself.

Note: These 4 cameras scored 3x,3x,5x.3x on the reduction from 34 to 28 so for the next set to be wildly divergent is worthy of an explanation. Even though the Sony and the Samsung are roughly equiv cameras hardware wise, to simpify the analysis lets just talk about the Sony by itself for a moment. On the first q setting of 34pass it had a 5x reduction, on the second a 1.2x reduction. Thats quite a difference!

Of course it could be explained by the scene complexity somehow, but then why wouldn't the Samsung also show dramatically less gains at the lower q? Instead we see dramatically higher gains after the lowering of q!.

Remember that although the administration of q thru a particular vendor is quirky and can't be relied upon, the h.264 definition of q is very formal and has to do with combining progressively more spatial frequencies. There are of course many other parameters invlolved in the actual bandwidth savings, like macro block size, and variable q settings between I and b,p frames, but assuming a motionless scene alot of these fall away and q should dominate, one would think.

As for your "strong likelyhood that decreasing compression will increase bandwidth" I would say its only trivially true: you would have been hard pressed with any camera on any scene on any day that would show an increase in compression level causing an increase in bandwidth.

Camera selection for the bandwidth section was random...

"Call it in the air, Heads I take the top-half of the chart, Tails I take the bottom. Tails it is..." :)

As far as the "plethora of variables" reason for only showing 4 cameras, goes, weren't all 8 cameras already set up at varying q levels for the Camera Compression Scales part of the test? It would seem you would just have to measure the bandwidth at that point, no?

In any event lest you misinterpret my musings as an attack, its not. I'm just trying to understand what's causing the data points to diverge. If the variance is inline with your expectations, then I will adjust mine. Thanks.

Login to read this IPVM report.
Why do I need to log in?
IPVM conducts unique testing and research funded by member's payments enabling us to offer the most independent, accurate and in-depth information.

Related Reports on Bandwidth

VLANs for Video Surveillance Tutorial on Sep 26, 2016
Many people confidently say to 'use VLANs' as an answer to IP video networking problems and as a way to signal expertise. But how should VLANs be...
How to Measure Video Quality / Compression Levels on Sep 16, 2016
Two cameras have the same resolution, frame rate and scene monitored but camera A consumes half the bandwidth than camera B. Is Camera A better?...
Camera Course September 2016 on Sep 15, 2016
This is the only independent surveillance camera course, based on in-depth product and technology testing. Lots of manufacturer training exists...
Milestone VMS Adds H.265, SVQR, RAM Video Optimization on Sep 09, 2016
Milestone is rolling out enhancements to XProtect to support H.265, enhanced edge recording functionality, and potentially allow users to reduce...
Hikvision 4K Camera Tested on Sep 09, 2016
Hikvision is the most common choice for low price entry level products but they are also competing with low light models, smart CODECs, WDR...
Pelco Optera 270° Camera Tested on Sep 06, 2016
Multi-imager cameras are typically 180° or 360°. Pelco has released a fixed 270° versions of their Optera intended to cover exterior building...
IP Networking Course on Jul 14, 2016
The reason for its popularity is that it is the only networking course designed specifically for video surveillance professionals plus it includes...
Axis Chinese Competitor Companion Dome V Tested on Jun 29, 2016
A $169 1080p dome from Axis? The Companion Dome V is, by far, Axis' lowest cost 1080p dome, a full $100 less than even their new M3045-V. Every...
Axis Launches Next Gen Zipstream on Jun 21, 2016
Smart codecs are one of the most important emerging surveillance technology trends. Axis was the first, launching Zipstream in Spring 2015 but was...
Selling 4K Cameras on Jun 21, 2016
This is a first in a new series to help integrators and manufacturers do a better job of accurately and fairly selling for and against various...

Most Recent Industry Reports

Camera Calculator Class and IPVM Member Orientation October 2016 on Sep 28, 2016
Members, learn how to better design video surveillance systems and get the most out of your IPVM memberships with 2 upcoming live classes. Both...
Axis 4MP Camera Tested (M3046-V) on Sep 28, 2016
Axis has brought 4MP to its camera line in the new M3046-V, the highest resolution model in their revamped M30 series. We bought and tested this...
Hiring Spree At Aimetis 6 Months After Being Acquired on Sep 28, 2016
Aimetis was acquired in April 2016, and is now expanding almost all of their departments, hiring employees from Axis and other industry...
Camio Natural Language Processing Tested on Sep 27, 2016
The ex-Googler led team from Camio has advanced its video monitoring offering to include natural language processing. Camio ingests video,...
Access Door Controller Configuration Guide on Sep 27, 2016
Properly configuring access control door controllers is key to a professional system. These devices have fundamental settings that must be...
Hacked Dahua Cameras Drive Massive Cyber Attack on Sep 27, 2016
Cyber attacks are accelerating and IP cameras are behind many of them. Worse, last week, a 'massive' attack was carried out using numerous Dahua...
Axis Secretly Paid Anixter Sales People To Push Axis NVRs on Sep 26, 2016
Internal Axis communication shows how Axis paid Anixter and Tri-Ed sales people with secret bonuses to push Axis NVRs. In this report, we examine...
VLANs for Video Surveillance Tutorial on Sep 26, 2016
Many people confidently say to 'use VLANs' as an answer to IP video networking problems and as a way to signal expertise. But how should VLANs be...
Ambarella CEO Admits H.265 and 4K Not Popular on Sep 26, 2016
Ambarella is the main chip provider for high-end surveillance cameras driving higher resolution and new CODECs. While Ambarella has been pushing...

The world's leading video surveillance information source, IPVM provides the best reporting, testing and training for 10,000+ members globally. Dedicated to independent and objective information, we uniquely refuse any and all advertisements, sponsorship and consulting from manufacturers.

About | FAQ | Contact