I don't care too much about color but if you look at the grass the 2MP does a LOT better. It's not even that there is much noise (there is some) but it's just plain fuzzy. Downsampling it reduces a bit of noise but it makes the image even fuzzier. That's kind of a surprising result. It's almost as if something is wrong. Well, I will see when I get that 8MP cam, not that I have something to compare it to, though...
4MP vs 8MP (IPC-T5442T-ZE vs IPC-T5842T-ZE)
- Thread starter rfj
- Start date
It isn't a surprising result once you realize that the optics just are not getting enough light if you downrez a low light image.
It is the same thing when people try to digital zoom and wonder why it turns all blurry and why they can't see plates like we do in the movies and TV...
Light is everything with these cameras. If you have enough light to begin with, then there is zero reason to consider even downrezing unless it is to lower bandwidth.
I can assure you if I pushed the 4MP to get brighter with higher gain, forced it into color, etc., you would see a lot of noise and it would get carried over to the downrez and probably result in even more soft blurry mess as it deals with the extra noise.
I bought the 4MP PTZ first and was disappointed in the performance and thought "Hey I will just downrez it to 2MP" and didn't see the results I was expecting. So I asked questions here and learned what I am sharing regarding the "pixel screen" and how downrezing doesn't magically let the optics see more light compared to the native lower MP counterpart.
Make sure when you test this that you are not on auto settings. The results would be a lot different if I ran it on default/auto settings because the shutter would be slowed down a lot to let a lot of light in. In that instance because the camera in auto/default settings is set up to favor a nice bright static image, it would result in a better looking downrez, but then in both instances it would be a blur.
If you really want to downrez, you probably need to feed the native resolution into some high powered computer program that can better account and adjust for a downrez than what is capable in these cameras. That software may be more expensive than the camera LOL.
Once something happens, you will wish you had at least one camera in color...
Infrared and B/W do nothing for color identification. Black pants can look white.
It is the same thing when people try to digital zoom and wonder why it turns all blurry and why they can't see plates like we do in the movies and TV...
Light is everything with these cameras. If you have enough light to begin with, then there is zero reason to consider even downrezing unless it is to lower bandwidth.
I can assure you if I pushed the 4MP to get brighter with higher gain, forced it into color, etc., you would see a lot of noise and it would get carried over to the downrez and probably result in even more soft blurry mess as it deals with the extra noise.
I bought the 4MP PTZ first and was disappointed in the performance and thought "Hey I will just downrez it to 2MP" and didn't see the results I was expecting. So I asked questions here and learned what I am sharing regarding the "pixel screen" and how downrezing doesn't magically let the optics see more light compared to the native lower MP counterpart.
Make sure when you test this that you are not on auto settings. The results would be a lot different if I ran it on default/auto settings because the shutter would be slowed down a lot to let a lot of light in. In that instance because the camera in auto/default settings is set up to favor a nice bright static image, it would result in a better looking downrez, but then in both instances it would be a blur.
If you really want to downrez, you probably need to feed the native resolution into some high powered computer program that can better account and adjust for a downrez than what is capable in these cameras. That software may be more expensive than the camera LOL.
Once something happens, you will wish you had at least one camera in color...
Infrared and B/W do nothing for color identification. Black pants can look white.
I really need to find some specs for those sensors, i.e. I want to know what the fill factor is for a 2MP camera vs a 4MP camera vs an 8MP camera on the same sensor size. Not that it matters for the camera I bought (maybe it was a mistake) but I just try to understand where this issue is coming from. Downsampling is pretty simple (ok, it can get complicated if you want to be sophisticated) and I think a camera can handle it just fine. No powerful computer is needed. Plus by the time the image gets to the computer it is heavily processed anyways, so too late.
I wouldn't want all cams on B/W but I have other overlapping cams that almost always record in color. Hence, I can get the color of cars/people from those other cams.
I wouldn't want all cams on B/W but I have other overlapping cams that almost always record in color. Hence, I can get the color of cars/people from those other cams.
This thread may help make it clearer, or make it more muddy LOL
ipcamtalk.com
Image Sensor Size & Pixel Size Comparison Calculations
I was recently comparing several cameras for a location I was interested in, and I wanted to assess light sensitivity beyond what is just listed in the spec sheet. I did a rough calculation and thought others may find it interesting. I calculated the sensor size using the number of vertical...
- Dahua 1/1.8" 4MP - 6.92mm x 12.30mm, ~85 sq mm area. 0.000021 sqmm / pixel
- Dahua 1/1.8" 8MP - 6.92mm x 12.30mm, ~85 sq mm area. 0.000011 sqmm / pixel
Last edited:
If those numbers are correct then I am back to square one. It means that the 8MP cam has a pixel size that is half of a 4MP cam (I suspected it might be much less than 50% because of the readout lines, i.e. ignore all the grid lines). Hence, if you downsample an 8MP cam to 4MP you should get very similar results (almost indistinguishable).
4mp or 4K based on your using environment, low light always 4mp is better, but the 5842 right now make the difference bit smaller than before old 4K IR ones.
I give up trying to explain it lol. At this point I think you are simply trying to justify to yourself your purchase lol (been there before as I explain below on this very topic lol).
You continue to fail to realize that under low light conditions that 8MP camera sensor is seeing a lot less light than the native 4MP on the same size sensor. The additional screen filter doesn't just disappear when you downrez and it still starts with the native resolution receiving a lot less light. It simply doesn't process a downtez the way you think it does.
Many of us have been there done that and proved to ourselves under low light conditions it doesn't work.
I thought the same thing once upon a time as well but I trust what @Wildcat_1 says as he tears apart these firmwares and provides tweaks and suggestions to Dahua on how to improve the firmware.
Well actually I didn't listen to him lol and went ahead and bought the 4MP because I was too anxious for an auto tracking PTZ LOL. I learned then with my own money that downrez to 2MP didn't improve the image when I was trying to justify my purchase to myself lol.
He remoted in one night to help me tweak the camera when I was still a NOOB here and the first thing he said was I need more light, even when running with infrared he suggested more visible light.
Eventually the 2MP came in stock so I got one and at night with low light it is no comparison - the lower MP will beat the higher MP on the same size sensor all night long. I still have the 4MP but have repurposed it. It is why many here say chase sensor size not MP.
Like I said if it were that simple, then everyone would chase MP and then just downrez if there wasn't enough light for good enough results, but it doesn't work that way so a lower MP camera is the better choice...
If you had an 8 cylinder vehicle and remove 4 of the spark plugs, would it run comparable to 4 cylinder vehicle? No it would perform worse than a native 4 cylinder.
Here was when wildcat schooled me lol:
"No, unfortunately thats a misconception that people have. In other words buying a 4K, running at 2MP or 4MP but running at 2MP for chosen resolution doesn't affect / improve the overall performance or light sensitivity of the camera at all. The sensor optic and MP pairing does not change due to your selection of resolution is probably the easier way to day it
"
ipcamtalk.com
While I am sure sensors and firmware have improved since the PTZ came out and you might experience better results than I did, the fact still remains that you can't defy physics and in low light conditions the lower MP is the better choice when compared to a higher MP on the same size sensor.
You continue to fail to realize that under low light conditions that 8MP camera sensor is seeing a lot less light than the native 4MP on the same size sensor. The additional screen filter doesn't just disappear when you downrez and it still starts with the native resolution receiving a lot less light. It simply doesn't process a downtez the way you think it does.
Many of us have been there done that and proved to ourselves under low light conditions it doesn't work.
I thought the same thing once upon a time as well but I trust what @Wildcat_1 says as he tears apart these firmwares and provides tweaks and suggestions to Dahua on how to improve the firmware.
Well actually I didn't listen to him lol and went ahead and bought the 4MP because I was too anxious for an auto tracking PTZ LOL. I learned then with my own money that downrez to 2MP didn't improve the image when I was trying to justify my purchase to myself lol.
He remoted in one night to help me tweak the camera when I was still a NOOB here and the first thing he said was I need more light, even when running with infrared he suggested more visible light.
Eventually the 2MP came in stock so I got one and at night with low light it is no comparison - the lower MP will beat the higher MP on the same size sensor all night long. I still have the 4MP but have repurposed it. It is why many here say chase sensor size not MP.
Like I said if it were that simple, then everyone would chase MP and then just downrez if there wasn't enough light for good enough results, but it doesn't work that way so a lower MP camera is the better choice...
If you had an 8 cylinder vehicle and remove 4 of the spark plugs, would it run comparable to 4 cylinder vehicle? No it would perform worse than a native 4 cylinder.
Here was when wildcat schooled me lol:
"No, unfortunately thats a misconception that people have. In other words buying a 4K, running at 2MP or 4MP but running at 2MP for chosen resolution doesn't affect / improve the overall performance or light sensitivity of the camera at all. The sensor optic and MP pairing does not change due to your selection of resolution is probably the easier way to day it
"Review - Dahua SD49225XA-HNR 2MP 25x Starlight + IR PTZ AI Camera with Deep IVS & SMD Plus
@wittaj No, unfortunately thats a misconception that people have. In other words buying a 4K, running at 2MP or 4MP but running at 2MP for chosen resolution doesn't affect / improve the overall performance or light sensitivity of the camera at all. The sensor optic and MP pairing does not change...
While I am sure sensors and firmware have improved since the PTZ came out and you might experience better results than I did, the fact still remains that you can't defy physics and in low light conditions the lower MP is the better choice when compared to a higher MP on the same size sensor.
Last edited:
IAmATeaf
Known around here
The key info here is that no matter what resolution you configure within the cam it will always capture in its native resolution.
The idea that dropping the resolution down will alter the way that it uses the sensor ie use a group of 4 as 1 just isn’t the way that they work. Yes you would have a similar amount of light falling on the grouped 4 but the cam will just downsize using software.
The idea that dropping the resolution down will alter the way that it uses the sensor ie use a group of 4 as 1 just isn’t the way that they work. Yes you would have a similar amount of light falling on the grouped 4 but the cam will just downsize using software.
Mike A.
Known around here
- May 6, 2017
- 4,147
- 6,948
Pixels can be "binned" together for some sensors. That's done in the case of some phone sensors. But that's not these.
Maybe the better way to look at it is along the lines of pixel aperture. You have X light over Y area for EACH pixel. That's your source signal so even though you have more of them, you're not starting from the same point when downscaling and simply dividing an equivalent source.
Maybe the better way to look at it is along the lines of pixel aperture. You have X light over Y area for EACH pixel. That's your source signal so even though you have more of them, you're not starting from the same point when downscaling and simply dividing an equivalent source.
Pixels are like water buckets, except instead of getting filled with water over time, they get filled with light. With different resolutions across the same sensor size, the amount of overall light hitting the buckets doesn't change, but how full each bucket gets filled does change. Imagine if you had 64 "units" of light and a 4x4 grid of buckets capturing that light. Each bucket will get filled with 4 units of light. If that grid is now 8x8 buckets, each bucket is smaller and will only get filled with 1 unit of light each.
Pixels also have electronic noise that needs to be overcome by a signal (the light) to have a high enough signal-to-noise ratio for a pixel to know what color it should show. Pretend that the electronic noise fluctuates between 0.5 to 0.8 units, and you only have 1 unit of light. It's harder to tell what color that pixel should be because 50%-80% of the data that the pixel receives is from electronic noise. Now let's say it receives 4 units of light instead. Now the noise data is only 12%-20% of the pixel data. This is why when you record videos in the dark there's a ton of noise, because there's not enough signal compared to the random electronic noise for the pixel to tell what color it should be. It will randomly change colors due to the random electronic noise, which across enough pixels will show up as noise in the image.
So it's not about pixel count, it's about the signal to noise ratio. For the same sensor size, each pixel receives more signal at lower resolutions, and less signal at higher resolutions. Downsampling won't combine the four pixels with 1 unit of light into one pixel with 4 units of light because it's not additive; combining the data from 4 poor sources of data doesn't net you high quality data. However, using the additional data, you might be able to downsample 4 pixels into, say, the equivalent of 1.5 to 2 units worth of light.
This is a pretty simplistic explanation, but should give you an idea why more pixels doesn't mean better night time performance.
Last edited: