4MP vs 8MP (IPC-T5442T-ZE vs IPC-T5842T-ZE)

[rfj]

I am deciding on a new turret cam. I read the threads about the two cameras. They both have the same sensor size but one is 4MP and the other one is 8MP. I understand that in most cases you want to go with the lower MP to have less noise (unless you have a really well lit area where the 8MP cam still gives you good video streams). But even if the brightness isn't that great, why would you chose a 4MP over an 8MP cam? All cameras have different settings to go to lower resolutions. I understand this isn't done on the HW level (you can't change the pixel size) but in SW (combining pixels). But even if this is done after the AD conversion (so in software) you should still get a better SNR/lower noise by lowering the resolution.

I guess my question is, if you lower the resolution of a 8MP cam to 4MP is it still noticeably worse than a camera that natively is 4MP? If not and the price is the same then I guess the 8MP gives you more flexibility in case your needs/lights/etc changes.

 

[looktall]

It still has the same amount of light hitting the sensor.
Changing how it looks after the fact won't fix that.

 

[rfj]

It still has the same amount of light hitting the sensor.
Changing how it looks after the fact won't fix that.

Well, that is kind of my point. Both the 8MP and the 4MP have the same sensor size so both get the same amount of light. If you scale down the 8MP to 4MP then the only difference is that with the 4MP the light is converted to 4MP before the AD conversion and with the 8MP it is done after the AD conversion. I would think both should give pretty similar results, i.e. for any practical purposes it won't matter if you have a 4MP cam or an 8MP cam downscaled to 4MP (in the camera setting, not after recording).

 

[looktall]

The 8mp camera has twice as many pixels for the same amount of light.
Scaling the 8mp image down still has to start with an image that doesn't have enough light.

In the day time the difference in performance between an 8mp and a 4mp on the same size sensor would be practically indistinguishable.
At night time it would be unmistakable.
Unless you add more light for the 8mp.

 

[tigerwillow1]

I ran them side-by-side for a few days. The 5842 did a lot better than I expected in low light conditions, what I consider only slightly less sensitive than the 5442. Not even close to a significant difference. It makes me wonder if there's more than pixel size at play here, possibly improved technology with the 5842's sensor.

 

[wittaj]

Downrezing a camera does not work - It is still using the 8 million pixels - the camera doesn't change the "pixel resolution screen" on the camera when you go from 8MP to 4MP. The sensor still needs 2 times the light going from 8MP to 4MP, so the native 4MP camera will result in a better image at night. The firmware will make some algorithm attempt at downrezing it, but it could be a complete crap image or a somewhat usable image, but if there is a concern that the 8MP isn't performing or wouldn't perform well at night, then it is better to go with the 4MP.

I have a 4MP and 2MP on the same 1/2.8" sensor and the picture quality is quite different between the two and the 2MP kicks it's butt at night.

In most instances, you want to get a camera that will perform at your 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.

My 2MP cameras outperform my neighbors 4K (8MP) cameras....why....because they are both on the same size sensor.

When we had a thief come thru here and get into a lot of cars, the police couldn't use one video or photo from anyone's system but mine. Not even my other neighbors $1,300 8MP system provided useful info - the cams just didn't cut it at night.

My neighbor tried the "I will just downrez the 8MP to 2MP" and the image was a soft dark mess.

His system wasn't even a year old and after that event has started replacing with cameras purchased from Andy based on my recommendation and seeing my results. He is still shocked a 2MP camera performs better than his 4k cameras and he cannot figure out why downrezing from 8MP to 2MP doesn't work properly... It is all about the amount of light needed and getting the right camera for the right location and downrezing doesn't change the physics of the camera.

 

[rfj]

@wittaj that is quite interesting that downsizing from 8MP to 2MP created soft images (besides the fact that it's now a 2MP image). If this downscaling happens on(!) the camera then there shouldn't be much blurring and the noise should be reduced. That makes me wonder what the (Dahua) cameras really do when changing the resolution to something lower than the native resolution.

 

[wittaj]

Noise isn't reduced because these are all digital images to begin with. It is still working with the same image from the 8MP, so if the 8MP has noise and problems that one hopes to eliminate with a downrez, it won't as downrezing will have noise as well. Garbage in = garbage out.

Then depending on how it digitally processes the downrez it could accentuate the noise and make it worse.

It is why we say if you don't have enough light for the higher MP camera you are better off to go with the lower MP camera.

It is always better to go with native resolution versus a downrez situation.

 

[rfj]

That is the part I don't really understand. For simplicity, lets assume we have a 8MP camera and a 2MP camera each with the same sensor size. So the 8MP camera gets 1/4 of the light per pixel compared to the 2MP camera. That results in much more noise in the 8MP camera compared to the 2MP camera. But the 2MP camera is just averaging light over a bigger area compared to the 8MP camera. And the 8MP camera can make up for that (after AD conversion) by averaging out the noise. It's similar to capturing 100 crappy images and averaging them to get one good image (done a lot in astronomy and cellular phones in night mode).

 

[wittaj]

What you are failing to realize is there is more "pixel screen" material on the 8MP, so two cameras of different MP on the same size sensor will result in the higher MP having more opaque "screen" material that impacts how much light gets thru. So two different cams on same sensor means the lower MP will allow more light on the sensor.

Use a Window for example. Which one is clearer to look thru the one with a screen or one without? The more holes a screen is produced with, the more material there is as well.

These sensors are small and we need to get as much light to them as possible.

An analogy to try to understand why cameras need so much more light - let's look at an 8MP camera and this 8MP needs at least four times the amount of light as a 2MP for the same sensor. The sensor size is the same in each camera, but when you spread the "screen" of 8MP worth of pixel holes across the same sensor, it now has 4 times the holes, but also 4 times the "screen material" than the 2MP.

Kind of hard to explain, but lets try to use a window screen as an analogy - take a window where the opening is fixed - that is the sensor - you add a screen to it (that represents 2MP) and looking out through the screen is a little darker outside because of the screen material. Now replace that screen with one that has four times the amount of holes (now it represents 8MP) and it will be darker looking through it because (while the resolution would be better) there is a lot more screen material.

So if your house is like most where the top pane is glass with no screen and the bottom half is window with the screen and you see something outside during the daytime - do you sit down to look out through the screen or do you stand up to look a the object through the window with no screen?

And that is accentuated even more at night time. Look out your window with and without the screen and it will be darker looking through the screen than without it. If you are looking out your window to see the stars or the moon, do you look out the part of the window with the screen, or the upper portion without the screen material?

Now obviously as it relates to a camera, you need to balance the amount of pixel holes with the screen material - too few holes (and thus less screen material) and the resolution suffers, and too many holes (and thus more screen material) and the more light that is needed.

Your thought of astronomy or a phone taking 100 crappy images and averaging them together to produce one nice image is essentially what the backlight is doing, but then it results in motion blur, which is why we recommend not using backlight at night.

No doubt you can tweak the parameters of the 8MP to look as bright as the 4MP but then you are increasing the likelihood of motion blur.

 

[rfj]

I am not sure but I think I understand your "screen" point. But isn't the fill factor of those pixels in the range of >=95% (a bit more for 2MP cams vs 8MP cams because there are twice as many lines in each direction). So having a few more "screen lines" doesn't matter that much, I would think.

The example about averaging for astronomy and phones was not the best as this is averaging over time (rather than space) which introduced blur. The point I tried to make is that crappy image in doesn't mean crappy image out (yes, it's multiple images in and one image out which changes things). But in case, if a 8MP camera is downsampling to 2MP the image was taken at the same time. Hence, there is no motion blur. Just combining pixels and hence reducing noise.

 

[wittaj]

Well I guess you will just have to buy an 8MP and downrez and see for yourself compared to a native 4MP or 2MP lol.

Take two different screens of different size holes and hold them up to a window and the one with fewer holes will let more light in.

As an example let's look at the 49225 2MP and 49425 4MP PTZ camera on the same size 1/2.8" sensor. Many people bought the 4MP version thinking more MP was better. It didn't perform as well at night. It was too dark so they downrez to 2MP thinking it would look as good as the native 2MP version of the camera and it wasn't. Many people own both versions of that camera and will tell you the native 2MP performs better at night and downrezing the 4MP to 2MP was worse than the native 2MP.

I am one that owns both. Believe me if the 4MP downrezed to 2MP would result in better performance I would be doing so. But it didn't.

Downrezing doesn't improve the amount of light the sensor sees. Too many people think downrezing is the same as buying the lower MP version and it isn't because of this.

These sensors are small and need a lot of light. If you are starting with an image that doesn't have enough light to begin with downrezing doesn't make it brighter. Post processing a poorly lit 8MP image to downrez will still have the same noise and artifacts and depending on how poor the light is, it may make the noise worse with the additional processing.

The only reason one would downrez is to lower bandwidth or mistakenly assume that downrezing turns the camera into the lower MP counterpart.

Most of us here favor less motion blur and a native resolution. So if the processing to downrez worked like you suggest by combining pixels and multiple images to reduce noise, that will result in nice static image that will result in poor motion performance.

You can simulate that right now. Turn on backlight WDR at night and that is essentially doing what you are saying in taking multiple images and combining pixels to eliminate noise. The static image will look great and be brighter than without WDR. Now add some motion to it and it blurs.

Turn off noise reduction at night and look at what a mess the image is. These images are processed enough as it is, so adding another process to then downrez is usually going to add more artifacts and problems.

 

[wittaj]

This question is asked a lot LOL (heck I even asked it in a camera thread after I bought a 4MP and asked if downrezing to 2MP would be the same as a 2MP camera LOL). So you can either believe those of us that tried or spend the money and try for yourself LOL (just please don't burn Andy by purchasing the 8MP and then returning it when you see downrez doesn't work)

Dumb questions of the day: if 4mp cam needs more light than a 2mp cam with the same size sensor....

....is there a way to set the 4mp into 2mp pix mode so that the 4mp cam only needs the same amount of light? In other words, if you purchase said 4mp cam thinking it will do just fine in low light but it turns out you overestimated the amount of light available at night is there a way to...

ipcamtalk.com

2MP vs 4MP vs 8MP - day vs night ?

A 4MP needs about twice as much light as a 2MP? However, can a 4MP operate as a 2MP? IOW, would it make any sense to get a 4MP or 8MP and have it at full rez during the day and then downrez at night? Thanks,

ipcamtalk.com

2Mp or 4Mp? Yes, I've read the reviews :-)

I've read the reviews that are available, but still don't have enough info to make an educated decision. I need at least one varifocal turret IP camera. Looking at DS-2CD2H25FWD-IZS (2 Mp) and the DS-2CD2H45FWD-IZS. I'd be using this camera to determine the fixed focal length of the other...

ipcamtalk.com

Dropping resolution from 4MP to 2MP on SD1A404XB

I am anxiously awaiting the delivery of a SD1A404XB tomorrow. First PTZ so it's very exciting to have a new toy arriving. Having done some reading, it seems that Dahua is using the same (or at least the same size) sensor 1/2.8" as the 2MP model of the same camera (SD1A203T). This camera will...

ipcamtalk.com

When/Where to use 8MP-1/1.2 vs. 4MP-1/1.8?

I'm wondering if there are general guidelines for when 8MP-1/1.2 cameras would be smarter to use vs. 4MP-1/1.8 cameras? Is a 1/1.2 always better (i.e., would turning down the 8MP camera to 4MP yield a better image than a 4MP-1/1.8 camera? Thank you!

ipcamtalk.com

2MP v/s 8MP if resolution is configured for 1920x1080

If you had a IPC-HDW3241T-ZAS 2MP which has a native resolution of 1920x1080 and a IPC-HDW3841T-ZAS 8MP and manually configured the latter to 1920x1080 resolution, would there be any appreciable difference in image quality? Both models feature a 1/2.8 CMOS image sensor.

ipcamtalk.com

Hi There

Hi all my plan is to put up 4 outdoor cameras and 2 indoor, I got 4 4k Amcrest and 2 5mp Amcrest cameras, I bought by Amcrest because I didn't want to buy Chinese cameras that talk back to china, I only learned after the fact that most if not all cameras come from china with different branding...

ipcamtalk.com

 

[rfj]

Well, I bought the IPC-T5842T-ZE from Andy... It's a refurbished cam so probably someone returned it for a 4MP cam... I didn't even know Andy accepts returns. I should read his return policies (not that I am going to return it).

I read through the threads but I still don't quite understand why you can't get very similar results with a 8MP cam downsized to 4MP. There are techniques to "physically" combines smaller pixels to larger pixels. I am pretty sure Dahua or any other security camera vendor are not doing this. Since the resolution is a setting in the camera, I would assume the downsampling is done on the raw data coming from the camera, i.e. before any other processing. This should result in less noise. I don't quite understand why this wouldn't be the case. Does anybody know how security cameras handle downsampling (most importantly at what stage but also what algorithm, etc).

In any case, once I get that 8MP cam (my first 8MP cam) I will mount it and see what I get. I have a few 4MP cams but not the exact same model in 4MP so I can't do a direct comparison. But at least I will get some idea how that 8MP camera performs.

 

[wittaj]

Well I guess we are not explaining it well enough and you will find out soon enough with your own experience LOL. There is another member here (forgot who but hopefully they see this thread) that also had both of these cameras and did a side-by-side and the 5442 was better at night.

You have to realize who the target audience is for these kind of cameras. They are mostly commercial and retail establishment with more light than we have as a residential user.

We are using these cameras to their limits with the low light that most of us have.

Is their pixel shifting and pixel binning technologies to to combine multiple smaller pixels into a larger pixel to increase sensitivity/reduce noise - yes - but they are not being done at this level. You will see that in mobile cameras that are taking static images and 4K projectors, but there just isn't a need in these.

Why buy in some cases a more expensive camera and then downrez it? If that resulted in a better image all of us would be doing that....

But whether the processing to downrez is done before or after any other processing, the fact of the matter remains that the sensor that you are downrezing from 8MP to 4MP is getting less light on the actual sensor than a 4MP on the same sensor. In daylight it isn't as obvious as it will be at night.

Look at a window screen. The window opening does not change size, so that is the sensor. The screen material is the "pixel screen". Which one of these is going to let more light thru?




So if you are taking a 22 mesh screen and trying to get the same amount of brightness thru into the house as the 4 or 14 mesh screen, it will need A LOT more light.

A close comparison would be the 22 mesh screen is an 8MP and the 14 mesh is a 4MP. So if you have the same amount of light going thru a 14 mesh screen as you do a 22 mesh screen, it will be darker for the 22 mesh screen and any parameter adjustment you do to brighten it (gain, iris, brightness) will add more noise.

Even if the downrez from 8MP to 4MPhappens before any other processing, it is still starting with a darker image than a native 4MP camera so that will introduce more artifacts.

This is how outside sun shades work. The darker/more shade it provides, the more holes the shade has. More holes equals more screen material.

Folks will purchase a camera with the native resolution they want based on the available amount of light they have.

 

[Mike A.]

I ran them side-by-side for a few days. The 5842 did a lot better than I expected in low light conditions, what I consider only slightly less sensitive than the 5442. Not even close to a significant difference. It makes me wonder if there's more than pixel size at play here, possibly improved technology with the 5842's sensor.

I also have both. As @tigerwillow1 said above, the 5842 isn't quite as good but it's not all that far behind. Where you'll see the difference is at the margins in low light. The 5442 has the edge and in most cases I'd rather have it. Where you have more light/IR you might not notice it so much. Where I first wanted to use mine with little to no light I could never get it quite as good as two 5442s that I had in two places with the same views before it. Ended up using it somewhere else and I think that it's fine for there without having something that I'm directly comparing it to. Good cam, I wouldn't discourage anyone from buying it. Just not quite as good as the 5442 as far as low-light performance.

 

[IAmATeaf]

The way I understand it, with the sensors being the same size the actual pixel size will be smaller on the 8Mp cam.

Assuming the aperture is the same size then the amount of light falling on the sensor will be the same but that amount of light falling on each pixel will effectively be lower on the 8Mp as each pixel is much smaller.

Now comes the issue, as far as I know the cam will always capture in its native resolution and will then down sample down which is why the image appears softer which sort of explains why reducing the resolution doesn’t equal the same performance as the native 4Mp cam.

 

[rfj]

I think my error is in assuming a completely wrong fill rate. I was thinking of something like 95% or more (95% of the area is used to gather light and 5% is for reading the pixels). Google didn't tell me much but it seems for those tiny 1/1.8" sensors the fill rate might be more in the 1/3 (35% or so) range (1/3 of the are is used to record light and 2/3 is for reading pixels). These chips probably don't use lenses either (to channel light from the whole are to that small 35% active area). So putting another readout line through that area lowers the fill rate and light considerably more.

Well, I got the 5842 and will make the best out of it. The area I am going to use it is a bit dark because the neighbor's house is like 10' away so I keep the lights low. But there is a light there. And my wife always tells me our house is like Las Vegas. If you drive down our street you will understand why she is saying that... Never mind when Christmas comes around...

 

[Mike A.]

It's a good cam and at a refurb price well worth getting. You'll very likely be happy with it.

One oddity that I saw at least with the firmware that mine has is that SmartIR has way too much of an effect in darkening the overall scene. You could see it flash brighter and look great for a second and then it adjusts everything down to the point that you can't see much of anything at all no matter what else you do for settings. If you see the same, turn that off and it will be fine. SmartIR rarely works all that well on most Dahua cams anyway. You can balance that out pretty well with other settings in most cases.

 

[wittaj]

Here is a real world example from the 49225 and 49425 PTZ I mentioned earlier.

The 49225 is a 2MP PTZ on the 1/2.8" sensor. It deems it has enough light at a 1/60 shutter so it stays in color:



Here is the 49425, which is the same camera as the 49225 except is a 4MP on the same 1/2.8" sensor as the 49225.

Here is the first big issue you see with a double the resolution on the same size sensor - It deems that it does not have enough light at a 1/60 shutter so it goes to B/W with Infrared:



The first thing you notice is that downrezing the 4MP 49425 down to a 2MP doesn't result in the camera being able to run color like the native 2MP 49225 as the sensor still doesn't see enough light to run in color because the 4MP "pixel screen" simply isn't letting enough light get into the optics of the sensor.

That is a big deal with low light conditions and why you want to go with a native resolution and preferably a camera on the ideal MP/sensor ratio that is talked about here often. The native resolution may be able to be in color, but the higher resolution on the same size sensor probably won't with low light conditions.

Then in this case, you can see that the 4MP was struggling to even give definition compared to the 2MP. It is wet out so the rain reflections is wreaking havoc with the infrared and focus.

So when I downrezed but kept the bitrate the same, it still looks like a soft mess. Even if there is less noise in this instance, it still doesn't look as good as the native.



As always YMMV and I am sure the newer cameras are improved over when this camera came out, but even so I don't think we see a higher MP downrez come close to the performance of a native MP on the same size sensor because the higher MP just isn't letting enough light into the optics as the lower MP on the same size sensor.

And the less light that is available, the bigger the difference one will see. Personally, for me the difference between being able to run it in color versus B/W is enough of a reason to go with the the camera that is on the ideal MP/sensor ratio talked about here so much.