Ethan
You were not referring to other aspects, which I believe are important as well. Do you have any information on the bandwidth consumption differences between the 3 types, as well as the effective frame rate each camera produces?
Thanks
Ariel
10MP vs 5MP vs HD Shootout
Published Nov 20, 2013 05:00 AM
** ******* *** ***********. ******** ****** 1080p. *** *** ******** *** ** how **** ****** ** *** ** 10MP?*** **** ****** * ***** ****** than ***** ** ***** ***** ********** claim?
** **** ******, ** ***** **** results ** * **** ** **** shootout ******* *****, *** *** **** cameras. *** **** ****, ** ****** 3 ******* *** ******* **** ******* (3MP ******, *** ******, *** **** AV10115), ***** *** **** ***** *********** claims *** **** ********** *********. **** allowed ** ** ********* *** ********* across ************* *** ***** ****** ** the ********** *** ************* ** ******* resolution
Key ********* ********
*** ***** ******** *** ********* *** questions:
- *** **** **** ******* *** *** higher ********** ******* ******* ** **', 50', ***' **** ****?
- ***** ****** *** *** **** ** low ***** *** ** *** ****?
- ***** ****** *** *** **** ** strong ********* ****** - *** *** HD ****** ** *** ****** ********** 5MP ** **** ****?
- ** **** ***** **** ********** ***** fields ** **** (~**' *** *****), 5 *** ** ** ******* ********* no ********* **** ** ******* ** our **** ***** *** ******* ******** to *** *****..
- ********** ********** *** ** ~**', *** 1080p ****** ***** **** ******* ** both ******* *** ***** ***** *** 5 *** ** ** ******* ******* recognizable ******* ******* *** ***** **********.
- ******** *** *** ** ~***', **** the * *** ** ** ******* provide ******* ******* ** *** *******, while ** *** *********** * **** in *** ***** *****. **** **** provides *** ********** ** *** **** chart, **** ** **** *.
- ** ***** *****, ******* ******* *-* lux, ** * **' **** *** 5MP ****** ******** *** **** ******** between *** ***** *********** *** **********, delivering **** *** ******* ******* **** both *** ***** *** **** ******.
- ** **' ********** *** ** ***** light, ******* * ** ** ** cameras ******** *** ****** ** *** test *******, *** *** **** ********** chart *** ***** ******* (****** **** the ***** ****) ** *** *** camera. *** ***** ****** ******** ***** details ** *** *******, *** *** chart *** ****** *******.
- ** ***** ***** ** ~***' ***, the ***** ****** *** **** ** provide ********* ** *** *******, *** not *******. ******* *** * *** 10 ** ******* ******** *** ****** video ** **** *****.
- ** ********, * ***** ***** '***** low' ***** **** ******, ********* *** brighter ****** ****** *** **** ***** as **** ** ****** ** *** narrow ***.
- ** ***** **** ******, *** *****, 5MP, *** **** ******* *** *** provide usable ******* ** ******* ** ***** at *** *****. *** ***** ****** was **** ** ********* ****** *** chart, *** *** *** *******.
- *** ***** ***** ***** *** ***** 720p ****** ** **** ***** ********* brighter ******, ****** ******* *********** ******* due ** ********* *** ***** ******.
- *** ***** **** *** ****** ******* WDR ****** **** ** **** ***** and **** *****. *** **** ********* was *** ****'* ******* ** **** lower ***** ** *** **** ***** but **** *** ***** ****.
- *** **** ****** ******* *** ******** backlit **** ** *** *** ***** well, *** ****** **** *** *** counterpart. *******, ** *** **** **** next ** *** ******** ****, ******* camera ******** ******* ** *** *******.
- ******* *** ****** - *****
- ******* *** ****** - *****
- ******* **** ******* - *****
- ***** ***-**** - ********
**** *** *** *** ******** **** this ****:
*** **** *****
*** ***** ***** / ******* *** (3-8lx)
*** *** ***** (*** *** *****)
***
***************
** **** *****, ********** ***** *** light ** *** *********** *** * concern, ***** ** ***** * ****** choice **** ****** ***********. * *** 10 ** *******, ***** ********** ****** pixels *** ****, ****** **** ****** performance ** ***** ******.
** ****** *** ******, **** ** well *** ******** ***** ** *** parking ****, * *** ** ** resolution ****** ********** ****** **** ****** of **** (**'+). ** ***** ******, 1080p ** **** **** ****** ** deliver ********* *******.
** *** ******, **** *** ******* should ** ****. ***** *** **** model ** **** **** ********* ****** than *** *** ***********, **** **** outperformed ** *** **** *****-******** ***** model. *******, ***** **** *** ** generally *** ********* ** ******* ***** 3MP, ** ***** ***** **** ** multi-door ******* *****, **** *** ** beneficial *** ** *** ****** *** performance.
FOV ***********
** **** *** ********* *********** ******** on ** ******** ***, ** ***** light ******. ***** *** ******** ***** *** ****** ***** ******.
**** *****, ~***-*** ***
** * ********** ***** ** **** of ~**', *** * *** ** MP ******* ** *** ******* **** practical ******* **** *** ***** ******. Our ******* ** ************ ** *** three, **** *** ***** ******* ** line */* ** **** *** ***** camera.
** * **' ****, *******, *********** become ********. *** * *** ** MP ******* ******* **** **** ******** of *** *******, **** *** ***** remaining ******* **** ***** *****.
*******, ******** *** ***** ** **** to ***', **** ****** *** **** performance, **** ****** ******* ** *** subject, *** *** ***** *** ****** lines ** *** ***** ***********.
***** *****
** **** *****, ******* ******** ******** between *-* ***, ********* **********. ** * **' ********** ***, the *** ***** *** *** ***** camera ** ******* ** *** ******* and *****. *** *** ***** ****** produced **** ***** *** *********** ** this ***** *****, ********* **** *******. The **** ***** ******* ***** *** test *****, *** ** ******* ** the *******. *** *** ************ ** a **** *** ***** ********* ******, shown *** *********.
** * **' ****, *** *** camera ***** ***** **** ****** ** the **** *****, *** *** ******* is ****** ********** ** ***. *** 1080p ****** ***** **** ******* *** chart, ****** ******* *** *******, ***** the **** ***** ** ****** **** to *** *** **** *****.
*******, ** *** ****** ***, ***', the ***** ***** *** **** *** light ****** ***** ******* ********* ** our ******* *** *****. *** *** model ********* ***** ****** ** *** test *****, ***** *** **** ****** is ****** *****, **** **** ****** of *****.
****,
*******, ** * ***** **** *****, the ***** ****** ** **** ** detect *** **** *****, *** ******** no *******. **** * *** **** cameras **** *********** *****, **** ** usable *****.
** * **' ****, *** ***** camera ****** **** *** ***** ** all.
*** *******, ***** * ***' ****, none ** *** ************** ** ******* ****** *****, **** even *** **** *** ***** ****** only ********* ********* *** *****.
WDR ***********
** **** *** ***********, ** *** up *** ******* ** * ********* with * **** **** ******** ****, creating * *** ******** ** *** levels, **** ** **** *****:
*** ***** **** *** ****** *** the ***** **** ********* ** **** scene. *******, *** **** *****, *** to *** ****** **********, ********* **** in *** ****** ************ ** *** overhead ****, ** ****. ******* * nor **** ****** ******** ******* ** the **** **** **** **** ** the ****.
Interior ******* **********
******** *********** ** * *** ** MP ********** *** ** **** ** this ******* **** **** *** ****** ********* *****. ** **** ****, *** ************ both ***** *** ****, ***** *** best ******* ******* ********** *** ***** handling ** **** *****, ***** ** lux ** ******.
Test **********
*** ******* **** **** ** ******* settings *** **** ****, **** *** exception ** ********, ***** *** ********** at ****, *** *******, ***** *** turned ***.
*** ********* ******** ******** **** ****:
Comments (34)
AF
Ariel Frischoff
Nov 21, 2013
Kevin Bennett
Nov 21, 2013I, too, would be interested in the bandwidth comparison if one is available.
Ethan Ace
Nov 21, 2013The claimed frame rates are 21/14/7, respectively, for the 1080p, 5MP, and 10MP cameras. We found that 5 FPS was the max the 10MP camera would consistently output, though.
Bandwidth was tricky. Despite all three cameras coming from the same series from the same manufacturer, with all other settings standardized, and running at 5 FPS, there was not a clear upward trend in bandwidth as resolution went up. There are obvious sensor differences in sharpness and saturation which are at play here, and differing gain control on each of them which make it difficult to compare apples to apples. Because of all the above, we chose not to include it here.
Luis Carmona
Nov 21, 2013Great report! I like the by the numbers, to the point reports. Thanks.
JJ
Jeff Jones
Nov 21, 2013My guess is that the 10MP camera did better in your wide dynamic scene only because of the coincidental exposure strategy used. I would not take that result as an indicator of the camera to do well with other wide dynamic scenes. If you need a camera capable of properly imaging scenes with wide dynamic range, you need to use a camera designed for wide dynamic range imaging.
JH
John Honovich
Nov 21, 2013JJ
Jeff Jones
Nov 21, 2013The camera picked whatever exposure strategy it thought applied to the scene. That's what I meant by coincidental. Even though the cameras came from the same manufacturer, it doesn't mean that they necessarily apply the same exposure strategy to a given scene. Cameras are typically tuned independently, and often by different people. The 10MP camera could have simply "lucked out" with that particular scene.
JH
John Honovich
Nov 21, 2013We have done many tests over the years and found that more pixels has a 'WDR' like effect and it's not just a coincidence. (Go back to our first WDR test when we had a 5MP non WDR, true WDR 720p and Pixim - same pattern). Specifically, the greater pixels do help in areas that are not overexposed (i.e., the 10MP is better than the 5MP here, specifically because the greater pixels help). However, looking at both sides (dark area), clearly the 10MP (and 5MP) has problems across broad ranges of light.
That said, we agree that a true WDR camera is still better for WDR scenes overall.
JJ
Jeff Jones
Nov 21, 2013I think that your "WDR-like" effect is actually the effect of a higher MTF from the higher resolution camera. As you said, "...the greater pixels do help in areas that are not overexposed". WDR exposure strategies are supposed to prevent overexposure.
CD
Chris Dearing
Nov 23, 2013Jeff, are you saying that the 10 MP sensor's frequency response happened to have dips where the scene had crests and crests where it had dips?
Maybe Ethan had some calibration shots with an MTF chart?
C
BR
Brian Ring
Nov 21, 2013What a great set of pics. How far out is the subject in the 50' HFOV, approx?
CD
Chris Dearing
Nov 23, 2013
david tonsberg
Nov 21, 2013Maybe take this report a bit farther...since more pixels on the same size sensor means smaller pixels that require more light or does it?
The cost of a device increases for each change in size or does it?
Is using the right lense is very important? Where are all the details for the lens that you used for each camera? e.g. format, mount, focal length, F no, iris type, filter, etc. What are the costs for a lens @ 1MP, 3MP, 5MP or 10MP?
Why would anyone deploy a camera over 3MP or 1080p/1440p when lensing and storage costs go up significantly, or does it?
Ethan Ace
Nov 21, 2013We used Arecon't recommended lenses for each of the cameras.
The 1080p camera used an F1.2 3-8mm, $100-120 online.
5/10MP used F1.8 4.5-10mm. About $220 online. Note that this lens had the lowest F-stop of any of the recommended lenses.
So yes, there is a price increase in lensing.
david tonsberg
Nov 21, 2013The video experts in the photography and broadcast industry absolutely say the lens and camera are a system equivalently matched for performance. There for a 10MP camera and a 10MP lens are required to optimally acquire an image. Arecont has yet to reveal whether they have accurately done this in marketing to their customers.
Is IPVM testing MP cameras with a matched lens?
Lenses for cameras above 5MP are hardly easy to come by in C/CS formats much less in a variable focal length in any format from 1/3 to 1". Here is what I found in 1/2" format which the Arecont is 1/2.5" @ 5MP and 1/2.3" @ 10MP :
Space - nothing over 3MP with vari-focal
Fujinon DV4x12.5SR4A-SA1L is 5MP lens for about $387 (amazon) DC iris/ir corrected
Computar E3Z4518CS-MPIR is 5MP lens for about $ 243 (amazon) manual iris
Tamaron had 1/2.7" and 1/2" format MP vari-focal lenses however nothing over 3MP is defined on their web site.
Tokina TVR4518HDDCIR is 5MP 1/2" format with DC iris @F1.8. This is probably the lens sold by Arecont. $219 online @ B&H
Avigilon uses a EF type lens to open up the range of lensing options available in the higher MP range. These type of lenses have a wide range of costs from $600 to over $1500 depending on focal length and size.
So is the question of value versus function still a valid debate when the cost of the device and storage go up at a rate of 2X or more with the same rate in less performance when there is less/more light?
JH
John Honovich
Nov 21, 2013David, we used the Arecont approved lens - UHD45-10 which Arecont markets as being "Ultra HD Megapixel Vari-Focal Lenses Are Used In Conjunction With Arecont Vision 5MP And 10 MP MegaVideo Cameras To Deliver Maximum Performance And Image Quality."
I still wonder if that is truly a 'real' 10MP lens. That said, the issues for 10MP in this test all have to do with lighting (either low or harsh) - something that is more an imager / processing issue rather than primarily lens constrained.
BH
Bohan Huang
Nov 21, 2013Would using a F1.8 lens reduce the low light perfomance of the 1080p cam to levels similar to the 5mp (which used a F1.8)?
JH
John Honovich
Nov 21, 2013It's a false premise because 1080p cameras have widely available low F stop lenses where 5MP and 10MP ones do not.
Certainly a lower F stop helps but (1) I doubt it's all or even most of the difference and (2) it would be unfair and unrealistic to handicap the 1080p camera to match the deficiency of a 5MP or 10MP solution.
JH
John Honovich
Nov 21, 2013Btw, I do think one good counter would be to compare against a better 5MP camera - which I suspect the Axis P1357 would be. This is a potential future test.
For this one, the deciding factor is that we wanted to test a 10MP camera but Axis (and most major manufacturers do not offer them).
Ethan Ace
Nov 21, 2013Actually the Bosch Flexidome Micro 5MP that we've got in our upcoming test performed surprisingly well in low light. And by "surprisngly well", at least it produced an image instead of being all black. This is 1lx in the conference room:
I don't believe Bosch has any higher end cameras at that resolution, though. No boxes or domes.
AD
Andrew Del Biondo
Nov 21, 2013What bit rate was each camera set at?
JH
John Honovich
Nov 22, 2013We use VBR on tests, so the bit rate is allowed to fluctuate to keep the quality / quantization constant.
U
Undisclosed #1
Nov 22, 2013Are you using testing the AV3116 at 3MP or 1080p? The AV3116 is their 3MP WDR where as the AV2116 is their 1080p WDR. And if at 1080p, curious as to why you would step down the res instead of using the 1080p model out of the gate, or just test at the 3MP full res.
JH
John Honovich
Nov 22, 2013We ran tests internally at both using the AV3116v1. We showed / featured HD mainly because many people are caught up by HD specs (i.e., 1080p is 'true' HD).
As for image details in the test, 3MP vs 1080p is not going to make a big difference as almost all of the benefits of 3MP vs 1080p comes from the increased vertical coverage (1536 vs 1080). That said, in practice, I would typically recommend the 4:3 3MP over a 16:9 1080p (see our Aspect Ratio Shootout) simply to get the taller FoV.
DK
Dan Koz
Nov 25, 2013It would really help if you'd add moving targets into the analysis. Our biggest problem is the static 10mp view is great, but the minute someone or a car enters the frame moving it's always blurry and at low light it's totally useless.
JH
John Honovich
Nov 25, 2013Dan, thanks for raising this.
What's most likely happening there is that many camera manufacturers default to slow shutters (Axis is often 1/6s, I believe Arecont is 1/8s). That's what is causing the blur.
Here is a visual example of slow shutter/high blur vs normal shutter/no blur:
We run all tests at 'normal' 1/30s shutter which eliminates most blur and more importantly evens performance (since some manufacturers go for really slow shutters as a default and others go for more 'normal' ones).
This is also important when testing cameras because if 1 camera defaults shutter to 1/6s and the other to 1/30s, the former will get 5x more light, which will make a massive difference.
Does that help?
DK
Dan Koz
Nov 25, 2013That does although we’re stuck with Axis 1604 as the best overall camera running the shutter speed at 1/120 otherwise if someone starts running or we can’t read the license plates. Does anyone know how arecont got that night shots they show as demo somewhere in asia? I’d love to get that quality but they won’t say what they used.
DS
Danny Stamper
Dec 08, 2013Ethan and John,
I'm still struggling with the HFOV factor in your tests. If I look at the three boxes in the three variations of HFOV from 20' to 50' to 100' (with the red boxes around the subject) the shot is the same in all three but the subject is further away. What I mean about HFOV is that the same section of the cars on both the left and right side are showing identically. So since I think I've read that the distance "from" the camera to the target doesn't matter as much as does the HFOV, how are any of these shots different other than the distance to target?
JH
John Honovich
Dec 08, 2013I do not understand the premise of your question. The cameras are mounted side by side on the same pole, which is stationary for the entire series of tests. We then move the subject progressively farther from the camera. As the subject moves out, the FoV becomes wider (e.g., 20', 50', 100').
See our FoV Tutorial.
CD
Chris Dearing
Dec 08, 2013Danny I think I see where you are coming from...
The HFOV is dependent upon an object(s) of interest, i.e. 'the subject', being selected from the frame by the analyst.
So if the subject were the cars on the side then you would be correct in saying that the HFOV does not change between shots (~15 feet).
However I find that all other things being equal, I tend to go with the guy dead-center conspicously holding the chart...
Although I must say I have lost no small amount of sleep wondering if Ethan's doppleganger were also present in the shot, where would the HFOV go?
C
-Actually I'm just assuming that the person in the shot is Ethan, but I don't really know :)
ML
Miguel Lomeli
Nov 18, 2014Hi everybody
I have some question, in your experiencie with security cameras, megapixels and all entails The cameras 1.3 MP, 3 MP, 5 MP and 10MP
Wich is the place apropriate for the cam of 10MP, 5 MP, 3MP and 1.3MP?
Who and where install the differents cameras?
I made a list of area that each camera can able of surveillance
MP Area
Camera 1.3 MP----- SUrveillance of personal without detail (regardless the face) For example in the oficce
Camera 3 MP-------- Surveillance of entry personal and vehicular (with details, faces, license plates )
Camera 5 MP-------- Surveillance of Boulevard and Intersection (brand of car, model, color)
Camera 10 ---------- Surveillance Freeway or higway (brand of car, model, color)
I wait your comments!
JH
John Honovich
Nov 18, 2014Miguel,
It would be useful to assign each of your categories a width of scene / FoV. I think that's what you are roughly doing but that would make it more precise.
For example, a 10 meter wide scene might be fine with 720P but a 30 meter wide scene might benefit from 5mp, etc.
Keep in mind, 5MP and 10MP, even with IR, have little benefit at night over 720p / 1080p and are typically worse in WDR scenes.
ML
Miguel Lomeli
Nov 19, 2014Hi John
the cameras of 5 and 10 MP wich is the environment perfect for give the maxim performance?
I want to think that this cameras is perfect install in a mall, supermarket is correct?
JH
John Honovich
Nov 19, 2014To maximize 5MP or 10MP cameras, place them indoors with even 24/7 lighting.