That does not in itself mean a sharper image. It could just mean that the length of the blur/ghost images are shorter. I am hoping that it does improve the focus of the object.
WizColor 2.0 is coming (15 April)
- Thread starter steve1225
- Start date
CCTVCam
Known around here
- Sep 25, 2017
- 3,140
- 4,056
...and the penalties - a larger aperture generally = shallower DOF (not sure any detriment has been detected on here but then again tests were only intial ones, at least the ones I've seen), the tests on here do suggest AI is currently less than perfect at removing motion issues caused by too much gain / too slow a shutter in very low light scenarios. The positives on the flip side from the new cameras are the larger individual pixel size and slightly brighter picture.
Unfortunately, overall I don't see any huge improvement and what there is, is negated somewhat by the motion issues.
I still say, the answer lies in 1"+ sensor sizes, smaller apertures (counter intuitive), better quality of lens ie clarity of lens and uniformity, more compression options including having very low compression / interpollation options available to maintain the ability to zoom heavily into background / off centre areas to detect crime happening there.
As Andy says though, for now, the current cameras look like their our only choice. So I guess you're either happy with the marginal gains / compromises or need a replacement camera for a failed one, or you wait. I just wish personally CCTV tech didn't seem so far behind drones & camera tech. It seems a long time since the 4Kt came out, and nothing better has come along to replace it or address the flaws. By now I would have expected cameras to have progressed onto 1" or even beyond.
Unfortunately, overall I don't see any huge improvement and what there is, is negated somewhat by the motion issues.
I still say, the answer lies in 1"+ sensor sizes, smaller apertures (counter intuitive), better quality of lens ie clarity of lens and uniformity, more compression options including having very low compression / interpollation options available to maintain the ability to zoom heavily into background / off centre areas to detect crime happening there.
As Andy says though, for now, the current cameras look like their our only choice. So I guess you're either happy with the marginal gains / compromises or need a replacement camera for a failed one, or you wait. I just wish personally CCTV tech didn't seem so far behind drones & camera tech. It seems a long time since the 4Kt came out, and nothing better has come along to replace it or address the flaws. By now I would have expected cameras to have progressed onto 1" or even beyond.
Techie007L
Getting the hang of it
"Significantly more light-sensitive than the 5442-LED (including less noise and motion blur at dusk or when running white illumination at night), sharper than the Color4M-T (without all its contrast, detail, color, fractal/AI image issues), and on par with the Color4K-T (just lower resolution and better contrast management)" should have been enough for you to place the camera. I am not at liberty to share just any security footage from customers' sites, nor have I the time or interest to set up multiple cameras for a proper comparison here just to satisfy your curiosity (which I have to do because of dishonesty—y'all will just claim "See, look at all that motion blur!" when it's doing better than most other cameras, and on par with the Color4K-T—you'd have to be reminded how bad the 5442 is without its super bright IR LEDs). I have already run multiple comparisons when the camera first came out and am more than satisfied—in fact, I feel sorry to have to move old 5442 inventory when the Color4M-TL exists, cheaper, and better for most scenarios (the main exception being when white light from the camera is undesirable).
Just make sure to adjust its settings as follows to start with—its defaults out of the box provide a disappointing picture: Camera -> Image -> Style=Standard (Soft causes saturation issues, pushing slightly colored objects to gray), Saturation=55, Sharpness=20, Gamma=45; LDC -> LDC=Off (causes blurriness and resolution issues); Exposure -> 2D NR=20 (remember you'll probably need to reboot the camera to make this setting stick due to firmware bug on all Dahua cameras—it reverts as soon as you close the page otherwise, resulting in a fuzzy image). If you want the white LED to come on sooner, you can use the Manual exposure mode and limit the maximum Gain setting (settings of 50-70 work well). If you want to reduce motion blur, you can limit the maximum Shutter time (note that there is a tradeoff; as the shutter gets faster and the AI brightens the image, the 3D NR gets more aggressive). I also like to set Illuminator -> Illumination Overexposure Remover=Number Plate Priority to darken the nighttime image a bit (note that it will also cause the camera to flash in the presence of moving vehicles at night; while I don't care for this behavior, it does allow the camera to capture license plates very well).Mine is supposed to arrive tomorrow.
In Camera -> Encode, I like the following settings: Compression=H.265, Smart Codec=On (extends the keyframe interval a bit beyond the maximum of 150 to 200 for better storage efficiency), Dynamic Compression=Off (causes severe compression/blocking artifacts), Resolution=1440p (1520p results in interpolation which will cause grid-like sharpness artifacts and drop your max FPS), FPS=30 (more frames increases the chance of you having a clear one for a snapshot of a moving object), Bit Rate Type=VBR, Quality=5 (good starting point, but adjust according to your scene/quality/bitrate needs), Max Bit Rate=Max (6144 Kb/s). Add I Frame Interval=150, Smooth Stream=100/Max clear where applicable (i.e. Sub Stream 1, or if you turn off Smart Codec). If you don't like the delay switching to HD in Blue Iris and don't care about storage efficiency, you can turn off Smart Codec and lower I Frame Interval.
In Camera -> Audio, you'll get the clearest sound with Audio Encoding=PCM, Sampling Rate=Max (16000), Noise Filter=Off. I am being unnecessarily verbose here because in another thread where I was discussing camera settings, someone intentionally set other settings I did not specify (which should have been obvious to anyone with knowledge) to bad values just to prove their faulty point.
Here's a before/after pixel-peeping (i.e. cropped) example of the camera on a nearly full moon lit night showing the difference between the factory default settings and mine (above):
That gets the image into a much better place—significantly improved detail and colors. After that, motion blur is just a matter of illumination and shutter speed, and this camera already demolishes the competition there (except for the Color4K-T, which it's on par with). To quantify performance in the scene above with the 5442: The 5442 would be very dark (maybe 25%) with poor color, details all blurred out into darkness, and while a moving person above would have modest motion blur, they'd be completely pixelating and fading in and out on the 5442 every time they moved. This camera isn't magic, but I don't know anything that comes close at its price point.
Sadly your wall of texts won't gain much traction or respect or whatever word you choose to use without examples.
Reolink cameras on 1/1.8" sensors show specs comp/better than 5442 yet real life performance shows they do not perform was well. We all know the manufacturers play with their specs. Heck on paper Axis should be poor performing compared to many other cameras but in comparisons they are the same/slightly better, but they may be the only true specs.... Who knows, but folks here want to see examples, not specs and walls of texts.
Wow you created a nice bright static image, but we don't care about that. I can slow shutter way down and crank up gain and get a comparable still image with the 5442 or a $40 camera.
Further, I, as others here, find it hard to believe this cheaper camera has better H265 and VBR qualities than the more expensive cameras.
Every test I have done, as others here have also done, show worse performance with H265.
We don't need you to spend hours setting up comparisons. And I appreciate and get your excuse of not wanting to share client video, but for your own reputation sake, as well as for the benefit of this forum, throw one up real quick somewhere in low/no light and walk around, put a mask on your face, whatever, to give us some examples.
If you feel so bad about moving your old 5442 inventory, offer it up to members here at the price of the 4MTL
And of course we recognize you throw enough light on anything, even many crap cameras can perform. But most of us here are not working with excess light and need something that can work with onboard lighting.
Last edited:
looney2ns
IPCT Contributor
Yeah, no video examples, so all this text is a waste of your time and ours.
Still pictures prove nothing, nada.
Still pictures prove nothing, nada.
So VBR, 6144 bitrate, and auto/default exposure?
Mmmmmkay
Most of the debate and confusion in this whole conversation revolves around "Night Color".
Getting good blur free, pixelated-free color night video with a target in motion is difficult. It takes a metric shit ton of light
The 5442 isn't particularly great at it, but its better than its predecessors and with minimal (1000-2000 lumens) it can do very well within 15 ft or so.
Everyone that has joined here in the past 2 years is under the false impression that color at night with moving targets is somehow expected and easy. And that's what the companies like Dahua are pushing.
But we see most recently with the 54PRO series, that all of the "Wizcolor, AI-ISP, marketing terms dont make for a better quality night color image. They make a BRIGHTER image.
Brighter isn't necessarily better
I'll take a good crisp blur free B&W/IR image over a poor fuzzy blurry color one when it comes to ID.
If all you care about is BRIGHT and COLOR then by all means, slap a camera with white LEDs up, set it to Auto, and forget about it
The image detail quality will suck, but it will be BRIGHT and COLOR and “Smooth”
Light fixes most things
With enough white light, many cameras can perform well at night. But the amount of light needed will surprise you
All cameras need light. No getting around that.
At night, they all need WAY more than you think and WAY more than Dahua (or HiK) wants you to think.
I have 11 cameras. Only 3 have enough light to provide a good clean crisp blur free COLOR image of a person walking past at a brisk pace.
The painful truth is that few residential scenes will have enough white light to run Color and produce a great blur/pixelated free image
Distance also matters
If you are using a wide angle (2.8mm or 3.6mm for example) lens with moving human targets at 30ft+, the size of the person is such that you dont get enough detail to matter.
This is why Manufacturers like to show you demo clips of skylines 100's of meters away. The smaller the object the less detail so that image quality imperfections aren't as noticeable.
Zoom in to capture facial ID and its a very different situation
Lastly samples with moving objects tell the tale. Not static images with far away subjects
Both cameras are 50-55 ft from capture point
- The 5459 PV PRO is directly under a very weak oldtime streetlight.
- The 5442 is 120ft away in the opposite corner of the house and does not have the benefit of a streetlight
B&W IR
Color - both under the streetlight
Last edited:
CCTVCam
Known around here
- Sep 25, 2017
- 3,140
- 4,056
I agree. I could make my 4KT's look like noon on a suny day with 300 lumens of light if I wanted to.
The trick would be to increase gain and slow shutter. I run 4ms or 1/250th at night on 300 lumens of light!!! I could easily slow the shutter to maybe 1/25 and increase gain to say 70 & get super bright summers day pictures. The penalty would be any object not standing still would be blurry and ghost like. Which would I prefer? A birght summers day picture that's full of ghosts or a dark usable picture that's got people anyone could recognise?
Therein lies the issue Techie. Getting a bright picture at night isn't hard. Getting a bright picture that contains usable detail of people is. Simply look at the pictures from many of the leading consumer cams. The picture of the neighbourhood looks great at night. Then add a person. We have a whole thread on it at the top of the forum contain hundreds of examples of great neighbourhood shots with unidentifiable suspects. That's what high gain / slow shutter usually gets you.
The trick would be to increase gain and slow shutter. I run 4ms or 1/250th at night on 300 lumens of light!!! I could easily slow the shutter to maybe 1/25 and increase gain to say 70 & get super bright summers day pictures. The penalty would be any object not standing still would be blurry and ghost like. Which would I prefer? A birght summers day picture that's full of ghosts or a dark usable picture that's got people anyone could recognise?
Therein lies the issue Techie. Getting a bright picture at night isn't hard. Getting a bright picture that contains usable detail of people is. Simply look at the pictures from many of the leading consumer cams. The picture of the neighbourhood looks great at night. Then add a person. We have a whole thread on it at the top of the forum contain hundreds of examples of great neighbourhood shots with unidentifiable suspects. That's what high gain / slow shutter usually gets you.
@bigredfish - The "Headless Horseman" (aka golf cart driver) is such a classic ... you will be posting that a decade from now! ;-)
@bigredfish is working on a color lighting flowchart. (in Metric Shit tons)
It will hopefully make things clear to the IPcamtalk community.
it will be depicting the " Metric Shit ton" column on the left and the equivalent lumens across the bottom.
Then in the middle of the graph find your sensor size.
Plot your findings on the graph.
This will give you your Shit ton lighting requirements in both Lumens and shit tons.
Then go to the lighting department at Home Depot and have a store associate find you the correct bulb for your calculated Metric Shit ton requirement in Lumens.
Thank you for your attention to this matter.
here are some
ossible ratios.
It will hopefully make things clear to the IPcamtalk community.
it will be depicting the " Metric Shit ton" column on the left and the equivalent lumens across the bottom.
Then in the middle of the graph find your sensor size.
Plot your findings on the graph.
This will give you your Shit ton lighting requirements in both Lumens and shit tons.
Then go to the lighting department at Home Depot and have a store associate find you the correct bulb for your calculated Metric Shit ton requirement in Lumens.
Thank you for your attention to this matter.
here are some
ossible ratios.Excellent tool, been waiting for this for ages. Really helpful. Thanks for you good work.
To be fair I left out one important thing:
“Good enough”
I’m admittedly hard core when it comes to image quality. Frankly only a few of my cameras in certain scenes are imho “great”. Most are “pretty good”.
It’s partially about the camera but more about how it’s setup and the available light vs distance vs speed of target.
I’m retired so pissing away 4 hours 2-3 nights in a row tweaking settings is a form of entertainment. Often when I’m done I’m disgusted that it isn’t “perfect”. My goal being great forensic evidence. And honestly my efforts often fall short of that.
It’s the reason I don’t play golf.
Face it, We’re dealing with $200 cameras. They aren’t “the best” , they’re more like “pretty good”
And for many “pretty good” may be “Good enough”
I get that and have to remind myself that “Good enough” might just be all the homeowner wants. Enough to be able to make out that it’s a person vs a moose and with the bonus that that person was wearing a red shirt.
That’s perfectly valid and there’s nothing wrong with that.
Just be sure to make the distinction
“Good enough”
I’m admittedly hard core when it comes to image quality. Frankly only a few of my cameras in certain scenes are imho “great”. Most are “pretty good”.
It’s partially about the camera but more about how it’s setup and the available light vs distance vs speed of target.
I’m retired so pissing away 4 hours 2-3 nights in a row tweaking settings is a form of entertainment. Often when I’m done I’m disgusted that it isn’t “perfect”. My goal being great forensic evidence. And honestly my efforts often fall short of that.
It’s the reason I don’t play golf.
Face it, We’re dealing with $200 cameras. They aren’t “the best” , they’re more like “pretty good”
And for many “pretty good” may be “Good enough”
I get that and have to remind myself that “Good enough” might just be all the homeowner wants. Enough to be able to make out that it’s a person vs a moose and with the bonus that that person was wearing a red shirt.
That’s perfectly valid and there’s nothing wrong with that.
Just be sure to make the distinction
When your new website sucks so bad you have to put out the option to return to the old version
Sadly, many people, including everyone in the Reolink Reddit LOL, would think this image is good enough! The still image is pretty nice though.
But this forum isn't like a reddit sub.
For most of us, we want to get a camera that will perform with an object in motion at our location for the worse situation, which for most of us is at night when it is dark and there is little to no light. If a camera performs at night, it is easier to tweak settings to make it work during the day than it is the other way around.
I am not looking for a nice bright static image to enter in the photo contest at the county fair - my mobile phone will beat any of these in that category. I want the ability to freeze frame an object in motion.
Now for those that don't know any better, or for those that are using it as an OVERVIEW camera, then I suspect installing the 2.8 or 3.6mm fixed lens 4MTL up on a 2nd floor will be fine like the PRO line examples because nothing is going to be IDENTIFY quality at the distances stuff is at, so a brighter image for OVERVIEW is great and for that purpose/goal, then yes go with a cheaper camera than the 5442.
But in that critical IDENTIFY distance of 15 feet, I cannot see the lower priced 4MTL doing better than the higher priced 5442 and probably will perform closer to the PRO series we have seen.
Maybe someone soon will get one and post some pics in that critical 15 foot ID zone.
I know from many sources that there is big firmware upgrade in development for 54PRO.. Latest version is from 01 April.
there are many new firmware subversions visible in internal Dahua firmware website, but my Polish Dahua people can't download it - it is limited by permissions for some internal testing group.
I hope that they will fix many problems with 54PRO..
Chokolinho
Getting comfortable
Just to clarify @steve1225 since you said "01 April" ... are you "messing" with us ... or is this legit?
it's legit... please remember that firmware is from Chinese team - so they don't know Prima aprilis..
I saw the screenshots of that website with that beta firmware...
Chokolinho
Getting comfortable
tibimakai
Known around here
Maybe with the new AI stuff, there is no need for a good sensor, AI will generate a good image, from a lower grade sensor(lol).