Less sharpness in T54PRO-ZE (versus 5442-S3) possibly due to noise reduction being applied even if set to ZERO?

A

alekk

Getting comfortable
Aug 13, 2018
279
430
Republic of Boulder
[Spinning up a separate thread from the "main one" if anyone want to comment SPECIFICALLY about this topic]

So I have the been comparing the T54PRO-ZE (running the 0731 firmware) and the 5442-S3 for several days now ... and one things that keep "bugging" me is the T54PRO-ZE pictures just don't look quite as sharp as the 5442-S3 during the day ... and even sometimes at night, despite the exposure being "better" and less noisy of an image.

I'm wondering if this is because the the T54PRO-ZE is applying some (poorly done) noise reduction that reduces contrast/sharpness ... even when that is set to ZERO.

Below are two pictures taken at the same time (mid-afternoon, partly cloudy) with the (manual) parameters identical at 0-10msec (1/100s shutter), ZERO gain, and ZERO NR. Note that if I change the shutter speed to 0-5msec (i.e. twice as fast), the exposure remains the same ... so there is plenty of light here.

The 5442-S3 is consistently slightly darker (the T54PRO-ZE appears to apply some "boost", especially in the shadows) and (during the day), the 5442-S3 has deeper colors. The 5442-S3 also looks a bit SHARPER to me.

I think(?) what might be going on here is despite have ZERO gain and ZERO NR, some NR is being applied.

For instance, take a look at the road - it actually DOES look "pebbly" as seen in the 5442-S3 image ... but the T54PRO-ZE is much "smoother" which I'm guessing is because it is interpreting that as noise and smearing it. Ditto if you look at the pitted driveway. And look at my "color chart" in the middle of the driveway - there's more detail with less smoothing around them in the 5442-S3 versus the T54PRO-ZE.

Having said that, areas like the tree on the right side are look sharper, but the increased brightness may contribute to that.

BTW, two interesting objects to look at are the white mailbox (on my side of the street) and the street sign just to the left of it. The mailbox has some purple chromatic aberration on the top of it in the 5442-S3 image, but there is hardly any with the T54PRO-ZE. And notice the sign has a "halo" up&down it from the 5442-S3, but it's not there from the T54PRO-ZE. Unknown if some of that is being corrected optically or in the post-processing (intentionally or as a by-product of the NR).

I'm real happy with the improved low-light capability of the T54PRO-ZE ... but wondering if anyone else has seen this behavior as what appears to be maybe overly aggressive noise reduction ... and why the heck is it even being applied when it is set to ZERO?!?

I know that @steve1225 has commented about how the image processing is different for the Color-4's and I believe(?) has said the same about the T54PRO-ZE, so maybe that is happening here too - I don't own any of those cameras.



Below are the (10 MByte!) PNG's as taken by the web interface - the same thing is seen looking at video. And I see similar issues with NR (of similar values) is applied - it seem that there's just something a bit "off" with the T54PRO-ZE algorithm.

BTW, I would normally convert these to JPEGS with Photoshop (about 1/10 the size and hardly, hardly any difference), but wanted to show the raw data ... so sorry about the large size. Should be an option to generate JPEG's! ;-)

Zoom level on the 5442-S3 is 500 (I hit auto-focus several times and it hardly budged from 1580) and on the T54PRO-ZE, it's 550 and similarly, the focus is 1728.
The "(LOL) color chart" is about fifty feet (slant range) from the camera's. Per various posts, they are literally right on top of each other and I tried to get the framing similar.


5442-S3

2025_10_13_1504_5442-S3_10-0-0-0.png


T54PRO-ZE



2025_10_13_1504_T54PRO-ZE_10-0-0-0.png
 
Last edited:
  • Like
Reactions: Clark Griswald, Broachoski, JDreaming and 3 others
Well, this is a good one, thanks for posting. Agree with pretty much all your findings, the 54Pro almost looks soft in the centre and sharper at the edges, this is totally the opposite of what you would expect with a lens! almost hope you have a bad sample here honestly, or some misalignment with the lens.

Seems like detail is lacking most noticeably on the road, driveway and of course the items on the drive, texture lacking, very odd, looks as you say like NR which is being forced regardless. Almost like sharpness is 0 on the 54Pro and 50 on the S3! Exposure looks too high on the 54Pro also, similar maybe lacking in contrast? Like compression effect or similar somehow, can't make much sense of it honestly.

Does the 54Pro use the same sensor as the S3 do we know, sorry, lacking knowledge as out of touch the last 6-12 months on specs and sensors etc.
 
Last edited:
  • Like
Reactions: JDreaming and alekk
Please remember that 54IR and 54PRO are using different sensors from different companies which have totally different characteristics..

54IR are using sensors from Omnivision (typical sign is WDR at 140/150db level for 4Mpx plus only 4/8 Mpx variants).
54PRO / 34PRO / ColorVu are using sensors from SmartSense (typical sign is WDR at 120/130db on all models plus 4/6/8 Mpx variants).

Plus yes - the processing is very different. WizColor / AI-ISP behaves very different in most stages of video processing. And many camera settings works very differently.
 
  • Like
  • Love
Reactions: Clark Griswald, Broachoski, JDreaming and 2 others
steve1225 said:
Please remember that 54IR and 54PRO are using different sensors from different companies which have totally different characteristics..

54IR are using sensors from Omnivision (typical sign is WDR at 140/150db level for 4Mpx plus only 4/8 Mpx variants).
54PRO / 34PRO / ColorVu are using sensors from SmartSense (typical sign is WDR at 120/130db on all models plus 4/6/8 Mpx variants).

Plus yes - the processing is very different. WizColor / AI-ISP behaves very different in most stages of video processing. And many camera settings works very differently.
Click to expand...

Is it too early to tell if the 54Pro can produce better images, I sincerely hope so, or even equal for that matter! Now I feel like should have taken more advice from @steve1225 before jumping into the three of the 54Pro ZE! - appreciate your knowledge on this matter :clap:

I asked you know who and we get this....which does not inspire me with much confidence:

1760394665800.png

This would explain a lot, damn, this is less than ideal, hope I am wrong but have a bad feeling now after reading that sensor data. This would surely mean the 54Pro series will not be able to match the S3 image quality? What sensor does the 4K-T have out of interest, wondering if it is Sony, this would explain the impressive image quality for that model?
 
Last edited:
  • Like
Reactions: Clark Griswald, JDreaming, bigredfish and 1 other person
What FW are you running on the 54PRO?

Clearly the 5442 has a better image.

If I had to guess I'd say they purposely added some NR/Contrast/Sharpening tricks to the FW to counter the over-sharpened digital noise and overly bright inherent characteristics of this new sensor and processing.
 
Last edited:
  • Like
Reactions: Clark Griswald and JDreaming
Exactly the same thing I am finding in comparing the 54PRO to the 5442. The image of the 54PRO looks like it has Saran plastic wrap covering the lens. Slightly distorted, washed out colors, lack of sharpness edge to edge. I took mine down and replaced it until there is a firmware improvement. Currently running the 7.31.25 firmware (started with the 9.28.25 firmware).
 
  • Like
  • Sad
Reactions: alekk, JDreaming, bigredfish and 1 other person
tmxv4128 said:
Exactly the same thing I am finding in comparing the 54PRO to the 5442. The image of the 54PRO looks like it has Saran plastic wrap covering the lens. Slightly distorted, washed out colors, lack of sharpness edge to edge. I took mine down and replaced it until there is a firmware improvement. Currently running the 7.31.25 firmware (started with the 9.28.25 firmware).
Click to expand...
Oh dear, disappointing, wonder how much can be resolved on the firmware side. With Pro in the name, it needs to have some kind of performance advantage!
 
@bigredfish : Similar to @tmxv4128, I also am "Currently running the 7.31.25 firmware (started with the 9.28.25 firmware)."

@tmxv4128 : Depending on your workflow, you may be encountering this issue I documented on the 5442-S3's ... since that kinda gives an appearance of "Saran plastic wrap" ... which you'll "see" over the entire image, but more easily in very dark and very bright sections.

I put in HLC=1 (just ONE!) which seems to work-around the issue and I get a color space from 0-255 versus 16-235.
I'm guessing that results in something different in the image processing (@steve1225 said WDR can also trigger it) that is more similar to the 5442-S3's.

ADDED: here's a few images from that thread - again, this was when pulling JPEG's from the API versus taking a PNG snapshot ... and is a "color space" issue, not sharpness/NR ... but maybe something similar to you are seeing @tmxv4128 ?

Web-v-Web16-235-v-API-inside.gif


Histogram-WebPNG-vs-APIJPEG.gif
 
Last edited:
  • Like
Reactions: Clark Griswald, JDreaming, tmxv4128 and 2 others
The softness is obvious in at least 4 areas to my old eyes

1- The decorative stones in the foreground
2- The grass/yard
3- The street
4 - the flag and iron fence at far 1 o'clock

I suppose next its trying to play with sharpness/contrast/saturation/ and NR to see how close it can be made to match the 5442-S3
(Though I have a couple of pre-S3 5442's that I think will show the same.)

I have one of each of the new cameras coming, hopefully this week. I have a good idea on testing locations
 
Last edited:
  • Like
Reactions: Clark Griswald, JDreaming and EMPIRETECANDY
alekk said:
Direct link to the release notes - sounds like hardly anything changed between the ".1" version (dated 7/31/2025 ... and what my camera is running, but from firmware provided by @EMPIRETECANDY ) and the .2 (dated 8/27/2025) ... and nothing appears related to what I'm asking about here.
Click to expand...

They frequently dont even mention tweeks to FW with regard to image quality
 
  • Like
Reactions: JDreaming
You could try ROI on the soft areas. That could possibly isolate the difference between the sensor and the processing.
 
  • Like
Reactions: JDreaming
CamCrazy said:
Is it too early to tell if the 54Pro can produce better images, I sincerely hope so, or even equal for that matter! Now I feel like should have taken more advice from @steve1225 before jumping into the three of the 54Pro ZE! - appreciate your knowledge on this matter :clap:

I asked you know who and we get this....which does not inspire me with much confidence:

View attachment 229838

This would explain a lot, damn, this is less than ideal, hope I am wrong but have a bad feeling now after reading that sensor data. This would surely mean the 54Pro series will not be able to match the S3 image quality? What sensor does the 4K-T have out of interest, wondering if it is Sony, this would explain the impressive image quality for that model?
Click to expand...
Looks like another example of enshittification.
 
  • Haha
  • Like
Reactions: Clark Griswald, JDreaming and bigredfish
alekk said:
I appreciate you chiming in @steve1225 ... and yea, I'm kinda wondering of the copy I have is slightly "off" - have a second one on order.

Here's a GIF that flips between the two pictures every second (hopefully will show on the forum software - if not, download) where you can more easily see the difference.

View attachment 229839
Click to expand...

One more thing about differences - each camera have different "internal default settings"...

54IR-S3 is very well know from from over sharpening the image at default (50) settings.. it's how it was tuned by Dahua engineers...

54PRO have very soft default settings..Sharpening algorithm is also very different. Probably Dahua engineers tuned this like that because this camera use non stop HDR algorithm which also works as a sharpening tools (but using totally different algorithm).

They didn't wanted to sharp to much like they did in past on Full Color cams which in theory are parents of WizColor technology. I hated both Color4K and Color4K-180 cams for that - image from those cams always looked for me very artificial.

Plus there are compressions benefits. Textures like grass / trees / walls / roads are not important in security / CCTV way. They only use compression bandwidth (a lot!) and put strain on video compressor which must decide what is important (people / faces / vehicles) and what is not. Look for long list of compression problems on Color4K.

Yes - you can increase sharpening settings.

We should compare how those cams records people / faces / clothes / vehicle details... Especially on movement. At day & night..

I'm a big fan of AI Codec. I use it almost on all cams on all installations. Mostly in VBR mode, which works like CBR with ability to peak bandwidth when needed (something happens in video scene). When video codec have ROI information from IVS (this is how AI codec works) - here and here are people / vehicles in video frame - it puts more bandwidths (ROI) for those area. At a cost not important things like grass / trees / walls / roads.

But this is how I use / configure cams as installer on normal installs for normal users. Not for pixel peepers :)
 
  • Like
Reactions: tigerwillow1, quiet.tea and CamCrazy
So far it seems the 54Pro needs more adjustment to begin dialling it in, from current feedback at least. Should get mine running before end of the week, then I can begin tweaking. Of course it is normally the moving objects which matter as @steve1225 says above, but of course static gives us or at least me a good benchmark to work from. Something tells me that I will have a few hours spent playing with settings on these Pro models!
 
Last edited:
  • Like
Reactions: bigredfish
alekk said:
@bigredfish : Similar to @tmxv4128, I also am "Currently running the 7.31.25 firmware (started with the 9.28.25 firmware)."

@tmxv4128 : Depending on your workflow, you may be encountering this issue I documented on the 5442-S3's ... since that kinda gives an appearance of "Saran plastic wrap" ... which you'll "see" over the entire image, but more easily in very dark and very bright sections.

I put in HLC=1 (just ONE!) which seems to work-around the issue and I get a color space from 0-255 versus 16-235.
I'm guessing that results in something different in the image processing (@steve1225 said WDR can also trigger it) that is more similar to the 5442-S3's.

ADDED: here's a few images from that thread - again, this was when pulling JPEG's from the API versus taking a PNG snapshot ... and is a "color space" issue, not sharpness/NR ... but maybe something similar to you are seeing @tmxv4128 ?

View attachment 229840


View attachment 229841
Click to expand...


Has anyone tried messing with Gamma?
 
  • Like
Reactions: alekk and duplo
You must log in or register to reply here.
Top Bottom