Image Quality Issues

[rfj]

I got an IPC-D54IR-ASE S3. Attached is a screenshot of about 1/4 size (1/2 length, 1/2 width) from BlueIris displaying the came using the full area of a 4k monitor. I understand the camera is 2688x1520 and we look at a quarter of the size so something like 1300x700. However, I am still a bit surprised that a lot of areas are quite blurry. For instance the floor (red rectangle). Also on the chair edge (red rectangle) there are sections that are completely blown out (I guess that is kind of reasonable if the sun hits it at the right angle). But you can also see a LOT of edge enhancement. See the curved red line on the chair.




As for the blurriness, I thought maybe my settings were not good enough. So I set a higher bitrate, set it to VBR with best quality, reduced the fps to 15 with an iFrame interval of 15. This didn't help, though. So am I doing something wrong or are my expectations just wrong? I have other Dahua cams that show no such blurriness and strong edge enhancements.

 

[MikeLud1]

I got an IPC-D54IR-ASE S3. Attached is a screenshot of about 1/4 size (1/2 length, 1/2 width) from BlueIris displaying the came using the full area of a 4k monitor. I understand the camera is 2688x1520 and we look at a quarter of the size so something like 1300x700. However, I am still a bit surprised that a lot of areas are quite blurry. For instance the floor (red rectangle). Also on the chair edge (red rectangle) there are sections that are completely blown out (I guess that is kind of reasonable if the sun hits it at the right angle). But you can also see a LOT of edge enhancement. See the curved red line on the chair.

View attachment 195205


As for the blurriness, I thought maybe my settings were not good enough. So I set a higher bitrate, set it to VBR with best quality, reduced the fps to 15 with an iFrame interval of 15. This didn't help, though. So am I doing something wrong or are my expectations just wrong? I have other Dahua cams that show no such blurriness and strong edge enhancements.

View attachment 195206

Try these settings also lower the sharpness to about 30

 

[bigredfish]

I believe thats the Dahua part# HDBW5442R-ASE

Network Cameras - Dahua International

Network Cameras

www.dahuasecurity.com

You didnt mention whether its the 2.8mm or 3.6 mm but basically they have a close focus distance of between 4-5ft. So anything inside of that may have some mild blur

Agree with @MikeLud1 above. Generally speaking VBR will by nature provide some macro blocking/bluriness until it ramps up the bitrate with motion in the scene. I never use it as the ramp up time is too slow.

Try CBR, 8192 as mentioned or even higher (I run mine at 10,240 and 4096 is too low for a 4MP camera), and backing off sharpness a bit (I run mine 40-45) should help. Most all 5442 models come out of the box over-sharpened.

A real full size full main stream image would be more helpful. Also make sure ROI is OFF

 

[wittaj]

+1 on all of above.

VBR means it will lower the bitrate when there is no motion. So it could make it look like a low quality substream.

One of my cameras I set to 8192 bitrate and VBR and the bitrate would drop below 100 when there was no movement for awhile.

 

[rfj]

Yes, I think it's the IPC-HDBW5442R-ASE. It's 2.8mm. I made the recommended changes but there isn't really any improvement. See the attachments. Note that I temporarily disabled the sub stream just to make sure I don't look at a sub stream. The camera isn't really for security but just to see where our dog is and it is used as a motion sensor. Still, to me it looks like a cheap toy camera, very much unlike the other Dahua cams I have. So I must be doing something wrong. I included the full view of the camera. The blurriness is where the red rectangle is (there are also other areas). The over sharpening is marked with red lines (chair, railing, TV). There are many more. The most extreme one is probably by the TV which doesn't have a white border.

 

[MikeLud1]

Yes, I think it's the IPC-HDBW5442R-ASE. It's 2.8mm. I made the recommended changes but there isn't really any improvement. See the attachments. Note that I temporarily disabled the sub stream just to make sure I don't look at a sub stream. The camera isn't really for security but just to see where our dog is and it is used as a motion sensor. Still, to me it looks like a cheap toy camera, very much unlike the other Dahua cams I have. So I must be doing something wrong. I included the full view of the camera. The blurriness is where the red rectangle is (there are also other areas). The over sharpening is marked with red lines (chair, railing, TV). There are many more. The most extreme one is probably by the TV which doesn't have a white border.

View attachment 195278
View attachment 195275

Change the Compression to H.264H

 

[bigredfish]

Beat me too it.

 

[rfj]

I changed it to H.264H but it makes no difference. I did improve the over sharpness by going to Setting -> Camera -> Image -> Image and put the sharpness down to 10 (was 50). There is still a white bar around the TV but otherwise it looks a lot better, though also more "blurry". In any case, I am fine with that and the images I include here are with this new setting.

Now one thing I realized is that as I switch between VBR and CBR the image quality pretty much doesn't change. However, the bitrate reported by BI drops a lot, down to like 180kBytes/s. That's like 1.3kbits/s. I understand that with CBR I get about 2180kBytes/s because that is what I requested. And I understand that the rate can drop when going to VBR. But why does the image quality not change? With the image quality not changing, it tells me that 2180kBytes/s CBR only contains about 180kBytes/s data.

CBR Image:


VBR Image:


CBR Setting:


VBR Setting:


CBR Byte Rate:


VBR Byte Rate:

 

[MikeLud1]

Change the I Frame Interval to 15

 

[mat200]

I changed it to H.264H but it makes no difference. I did improve the over sharpness by going to Setting -> Camera -> Image -> Image and put the sharpness down to 10 (was 50). There is still a white bar around the TV but otherwise it looks a lot better, though also more "blurry". In any case, I am fine with that and the images I include here are with this new setting.

Now one thing I realized is that as I switch between VBR and CBR the image quality pretty much doesn't change. However, the bitrate reported by BI drops a lot, down to like 180kBytes/s. That's like 1.3kbits/s. I understand that with CBR I get about 2180kBytes/s because that is what I requested. And I understand that the rate can drop when going to VBR. But why does the image quality not change? With the image quality not changing, it tells me that 2180kBytes/s CBR only contains about 180kBytes/s data.

CBR Image:
View attachment 195287

VBR Image:
View attachment 195282

CBR Setting:
View attachment 195285

VBR Setting:
View attachment 195284

CBR Byte Rate:
View attachment 195286

VBR Byte Rate:
View attachment 195283

Do look also at the dori section of the cliff notes .. there is an equation you can use to determine the pixel per foot you can expect ..

With a 2.8mm lens you will have a wide FOV and quickly lose pixel density once you are at the 100 ppf range .. which iirc for a 3.6mm 8mp camera is about 10-12 feet .. I would need to double check.

You can try a test with a coke can box or other household items with larger type printed to compare and contrast having them at various distances from you camera.

 

[dante555]

I changed it to H.264H but it makes no difference. I did improve the over sharpness by going to Setting -> Camera -> Image -> Image and put the sharpness down to 10 (was 50). There is still a white bar around the TV but otherwise it looks a lot better, though also more "blurry". In any case, I am fine with that and the images I include here are with this new setting.

Now one thing I realized is that as I switch between VBR and CBR the image quality pretty much doesn't change. However, the bitrate reported by BI drops a lot, down to like 180kBytes/s. That's like 1.3kbits/s. I understand that with CBR I get about 2180kBytes/s because that is what I requested. And I understand that the rate can drop when going to VBR. But why does the image quality not change? With the image quality not changing, it tells me that 2180kBytes/s CBR only contains about 180kBytes/s data.

CBR Image:
View attachment 195287

VBR Image:
View attachment 195282

CBR Setting:
View attachment 195285

VBR Setting:
View attachment 195284

CBR Byte Rate:
View attachment 195286

VBR Byte Rate:
View attachment 195283

Yes, it's quite unusual for the switch between VBR and CBR to not have a significant impact on image quality. It's likely related to how compression settings are managed and how the H.264H codec is implemented. It could be a limitation of the equipment or software being used.

 

[rfj]

I did try setting iframe at every 15th frame. At 15 frames per iframe you can see the image changing every 1s whereas with 30 frames you can see the image changing every 2s. So I know it's working. But there is no change in image quality. I can see that the bandwidth reported in BI changed from about 180KBytes/s for 30 frames per iFrame to about 350KBytes/s for 15 frames per iFrame. That's pretty much double which tells me most of the bandwidth is spent for the iFrame. It kind of makes sense because there is no movement so the difference from frame x to frame x + 1s and frame x to frame x + 2s is minimal. Basically just noise.

But why does the image quality not change when switching from VBR to CBR? Btw, I checked the switch between VBR and CBR both in BI and on the web interface of the camera, just to make sure it's not something BI does. But I see the same thing, i.e. no change in image quality when switching from VBR to CBR. I think that's the source of the problem, i.e. that the iFrame is too much compressed, even with the highest CBR setting of 17664.

A 2688x1520 frame should be 2688 x 1520 x 3 = 11970KBytes. In VBR with an iFrame interval of 15 frames the bandwidth is 180kBytes/s. So the image is compressed by at least a factor of 11970/180=66. At that compression ratio, I am not surprised to see what I am seeing. But why is this happening?

 

[flynreelow]

love that teak table. just sold one like that.

 

[bigredfish]

I really dont think its too bad and I personally would take sharpness back up to at least 30-35

Even though you have good light, inside is always darker and harder to get a great image than outside.

Have you tried taking the dome cover off? I say that because I noticed a nagging smudge for weeks and finally after a good rubbing with a wet paper towel, its gone.

 

[rfj]

love that teak table. just sold one like that.

It's starting to look bad. I am planning to sand it down and do a new coat. That white, round patch is a small area I sanded down just to see what it looks like after the lacquer is gone. I probably should have done it on the underside rather than on the table... And since you mentioned teak, the floor is teak.

 

[flynreelow]

It's starting to look bad. I am planning to sand it down and do a new coat. That white, round patch is a small area I sanded down just to see what it look like after the lacquer is gone. I probably should have done it on the underside rather than on the table...

yea, mine was starting to look like that and was gonna sand down and re finish with

https://www.amazon.com/gp/product/B001003MXG/ref=ppx_yo_dt_b_asin_title_o03_s00?ie=UTF8&th=1

do you have any markings on the bottom of the table? manufactured by who, and where? thailand?

 

[rfj]

No markings and frankly I am not sure it's solid wood. The underside has the same wood pattern as the top, which makes me think it could possibly be solid wood but who knows unless you cut into it which I am reluctant to do. We didn't purchase it. I do know the floor is 3/4" solid teak...

 

[bigredfish]

Your image is 980k
My 4MP 5442's avg 3.5 - 4MB

Plus your image is showing 3784x2140?
My 4MP images, like most, are 2688x1520

What are you recording to and what are you using to output the image? Thats an 8MP image from a 4MP camera

?

 

[bigredfish]

I did try setting iframe at every 15th frame. At 15 frames per iframe you can see the image changing every 1s whereas with 30 frames you can see the image changing every 2s. So I know it's working. But there is no change in image quality. I can see that the bandwidth reported in BI changed from about 180KBytes/s for 30 frames per iFrame to about 350KBytes/s for 15 frames per iFrame. That's pretty much double which tells me most of the bandwidth is spent for the iFrame. It kind of makes sense because there is no movement so the difference from frame x to frame x + 1s and frame x to frame x + 2s is minimal. Basically just noise.

But why does the image quality not change when switching from VBR to CBR? Btw, I checked the switch between VBR and CBR both in BI and on the web interface of the camera, just to make sure it's not something BI does. But I see the same thing, i.e. no change in image quality when switching from VBR to CBR. I think that's the source of the problem, i.e. that the iFrame is too much compressed, even with the highest CBR setting of 17664.

A 2688x1520 frame should be 2688 x 1520 x 3 = 11970KBytes. In VBR with an iFrame interval of 15 frames the bandwidth is 180kBytes/s. So the image is compressed by at least a factor of 11970/180=66. At that compression ratio, I am not surprised to see what I am seeing. But why is this happening?

See above, something aint right

 

[rfj]

Your image is 980k
My 4MP 5442's avg 3.5 - 4MB

Plus your image is showing 3784x2140?
My 4MP images, like most, are 2688x1520

What are you recording to and what are you using to output the image? Thats an 8MP image from a 4MP camera

?

View attachment 195310 View attachment 195309

Well, it's a jpg so it's compressed. But when I look at the images I sent on this forum, they pretty accurately represent what I see on my screen.