Resolution vs Compression Tested

I believe I am quite straight about the examples, thank you. My 'conflation' stems from the practical advise section:

If you choose to [make adjustments on your camera's compression levels??] do so, try a higher resolution camera with higher compression settings to save bandwidth with minimized quality reduction.

Is this to be understood as procurement advise only? Would you rather to say 'if you fancy you need a 1080P, you might be better off buying (ignoring price) a 3MP unit and setting the compression maximally?'

Reactions:

Truman HW

•Jan 17, 2018

In reply to Roger Paine

his assumption.

we dont think about YOU and your property incessantly.

...... and you'll be fine ;)

Reactions:

Stanislav Utochkin

•Nov 25, 2014

Firstly, the bandwidth is an amount of information contained in video stream. To get the best quality we should tend to high bandwidth. Having a limit of bandwidth we have to reduce the amount of information using the following ways:
1. Reducing frame rate
2. Reducing number of pixels
3. Using compression
Number of pixels reduces real resolution. Its influence is clear. Compression is more intelligent. Compression rejects resolution of low contrast details and fast motions (interframe compression) but keeps resolution of high contrast static details. Therefore compression doesn't decrease resolution on high contrast test charts. Low contrast details are not noticable on the screen in live watching, so they are not important in television and cinematograph. Compression is intended for television and cinematograph. However in security surveilance the low contrast details can be very important. Moreover we don't know beforehand which details will be important and in many cases scene illumination (and contrast) can vary.
Thus compression level should be choosen carefully. Scenes with low contrast details (low light) must have low compression. In other case these details will be lost.

Reactions:

Andrew Del Biondo

John, this report is very beneficial to better understand the relationship between resolution and compression settings and the impact upon a users perceived impression of picture quality and standards. However, the image samples generally relate to static targets. If you have a moving target and considerable scene change would there not be an increased risk of more compression artefacts at higher resolution with higher compression? Which would you expect to look better and have sharper stills, a 6 IPS recording at 720p resolution at 2Mbps CBR or a 6 IPS recording at 1080p resolution at 2Mbps CBR assuming it was a Hikvision or Dahua camera at the highest quality setting?

Reactions:

John Honovich

In reply to Andrew Del Biondo

IPVM

"Which would you expect to look better and have sharper stills, a 6 IPS recording at 720p resolution at 2Mbps CBR or a 6 IPS recording at 1080p resolution at 2Mbps CBR assuming it was a Hikvision or Dahua camera at the highest quality setting?"

If you are using CBR, you can not control the quality settings, by definition. The camera will do that automatically (i.e, as it hits / exceeds the bandwidth ceiling, it will increase the compression level to maintain the bit rate requirement).

As for 720p vs 1080p, I would expect the 1080p to deliver more details than the 720p, even if it had moderately more compression than the 720p in your 2Mb/s fixed bitrate scenario (i.e., the increased pixels would more than offset the moderately higher compression).

Reactions:

Jon Dillabaugh

Pro Focus LLC

Your scenario would need an even, well lit scene to be true John. In low lit areas, a typical Dahua or HikVision camera will well surpass the 2Mbps mark natively and would then become heavily compressed. From the article above, it seems like the 720p would suffer far less than the 1080p.

So I think the real question become where is the trade off from higher res/higher comp to lower res/lower comp a net zero vs a bias one way or the other. I would assume this is a very subjective test and also model/firmware specific.

Reactions:

(1)

John Honovich

In reply to Jon Dillabaugh

IPVM

We'll run a test of 720p vs 1080p in low light this morning.

Please don't speculate. If you think something should be tested, either ask us to test it or submit your own test results.

Btw, we tested 7 models and the patterns were clear across the models, so your assumption that this is 'very subjective' and 'model/firmware specific' will need more evidence to back it up.

Reactions:

Jon Dillabaugh

Pro Focus LLC

John,

We both have speculated here to a degree. My usage of Dahua and Hik cams have shown me that, generally speaking, it is hard to keep them anywhere near 2Mbps in darker scenes, IR or not. It's not to say it isn't possible to turn down sharpness and change other settings to keep the bitrate as low as you can, but under 2Mbps is a stretch. So, as per your testing has shown above, if you have a lower compressed 720p image vs a higher compressed 1080p image, the 720p may win out. However, you suggested the 1080p camera would provide more detail in your reply to Andrew above, which seems to cut against your own findings. So you can see the source of my confusion.

As far as the subjective comment, I was referring to the judgement we all have to make about which image looks better. We can measure many things, such as resolution, compression, etc., but I don't know of a quantitative factor for image quality. I think we would all most likely agree on our picks of favored images, but that just means we are being equally subjective.

Reactions:

John Honovich

In reply to Jon Dillabaugh

IPVM

"So, as per your testing has shown above, if you have a lower compressed 720p image vs a higher compressed 1080p image, the 720p may win out."

Please stop speculating. We are literally testing this right now. We'll post the results when finished.

Reactions:

John Honovich

In reply to Jon Dillabaugh

IPVM

"In low lit areas, a typical Dahua or HikVision camera will well surpass the 2Mbps mark natively"

As a point of reference, the typical Dahua or Hikvision camera has integrated IR. More than 60% of both Dahua and Hikvision models have integrated IR, verified by the 400+ models we track in the IP Camera Finder.

So we will test a Dahua and Hikvision IR camera here.

Reactions:

Andrew Del Biondo

Looking forward to the results of the tests. Ideally would appreciate tests being also performed with good lighting conditions and with high scene change. System designers are always asking me what bit rate they should set a camera at relative to resolution and recording frame rate to achieve high quality play back without compression artefacts. It would be great to establish some guidelines to follow for popular cameras such as Dahua and Hikvision.

Reactions:

John Honovich

In reply to Andrew Del Biondo

IPVM

"System designers are always asking me what bit rate they should set a camera at relative to resolution and recording frame rate to achieve high quality play back without compression artefacts."

They should not set bitrates (i.e. CBR). Use VBR to ensure that you do not get compression artifacts.

We have done many bandwidth tests (each camera tests comes with bandwidth stats at the bottom). The level of bandwidth depends on the camera model and the scene.

The findings of this test is about patterns, not specific bandwidth levels for individual models. The same thing for a test like Testing Bandwidth vs Low Light (e.g., this shows that when lights go down, bandwidth goes up significantly, but the absolute amounts will vary).

Reactions:

Andrew Del Biondo

John, most of the projects we get involved with for end users and integrators require continuous recording with a guaranteed recording duration. Dahua hardware is typically deployed at VBR with a capped bit rate setting and the highest quality setting (= 6), typically an IP-HDBW3202 camera may be configured at 6 IPS, 720p resolution, VBR capped at 1,920Kbps highest quality setting (= 6) or at 6 IPS, 1080p resolution, VBR capped at 3840Kbps highest quality setting (= 6). This method allows us to guarantee a minimum recording period as we can calculate the worst case storage requirements and our field testing suggests at these settings we hold picture quality and don't get compression artefacts. However, knowing the "sweet spot" per camera model for encoding settings for holding picture quality with complex changing scenes and guaranteeing recording duration and bandwidth usage is very valuable and any further advice is much appreciated. Dahua claim their newer Eco Savy series such as the IPC-HDBW5202 is more efficient with compression and can provide savings with bandwidth and storage. Have you or can you do comparisons between these different series cameras to confirm if their claims are correct or not without sacrificing picture quality?

Reactions:

Ethan Ace

We did some additional testing with integrated IR cameras this morning, on this topic, and as well as our Bandwidth vs. Low Light test which we've updated and added to our Bandwidth Guide.

In answer to specifics regarding Dahua and Hikvision 720p/1080p in low light:

1080p does not have a major advantage over 720p in low light, even uncapped. See this example below. Details are only very very slightly improved in the 1080p stream. Quantization on both of these streams was ~29 in Hikvision and ~24 in Dahua (for some reason Dahua quality "4" didn't result in Q28 as it normally has in other models, so it was a bit lower than Hikvision).

With a 2Mb/s VBR cap, the same is true. Quantization is higher in the 1080p stream than 720p (Q35 vs. Q33 in Hikvision, Q30 vs. Q29 in Dahua), details delivered are still only marginally different.

Reactions:

Jon Dillabaugh

•Dec 04, 2014

Pro Focus LLC

In reply to Ethan Ace

Ethan,

Thanks for taking time to test this out. That said, I do have a few questions if you don't mind me asking.

1) The uncapped cameras you used were which models?

2) Did you happen to note the bandwidth usage for the uncapped tests?

3) When you said the quantization increased from 29 to 30 for the Dahua cams, was that true for both 720p and 1080p? They both had the same quantization with the 2mbps cap?

Again, great work. It's much appreciated by members such as myself.

Reactions:

Jon Dillabaugh

•Dec 14, 2014

Pro Focus LLC

In reply to Ethan Ace

Sorry to dig this from the grave, but I was still hoping for a response from Ethan above.

Reactions:

Kevin Mundy

Stanford University

John,

I don't think Benros has enough pixels between his eyes there, I would not be doing my job if I did not tell you :)

Look familar?

Reactions:

Alexander Ryltsov

Hi John,

do you have a camera with controllable preprocessor (noise reduction step)? so, is it possible to play trade-off between compression and noise reduction, keeping the same visual quality or same bitate and monitoring bitrate or quality change?

my understanding of your results with resolution / compression / bitrate is that low resolution with low compression ratio can't provide better visual quality as it effectively wasting bitrate to encode noise. and hi-res/hi-compression has a side effect of aggressive noise reduction.

or we can turn this inside out - instead of reducing noise algorithmically it is possible to reduce (lab-only, not in real life) noise physically, by providing more light. is it possible to test, how this hi-res/hi-compression vs lo-res/lo-compression will behave if you increase illumination 2 or 4 times?

can you take a try with that?

Reactions:

John Honovich

In reply to Alexander Ryltsov

IPVM

"do you have a camera with controllable preprocessor (noise reduction step)?"

Do you mean a camera that allows us to adjust the DNR setting? e.g., see: Camera DNR (Digital Noise Reduction) Guide

Reactions:

Alexander Ryltsov