My seller has sent me the chinese firmware but it is very old: General_IPC-HX4(2)X2X-Themis_Chn_P_Stream3_V2.400.0000.4.R.20150321
Dahua 4mp, 6mm turrets (IPC-HDW4421C)
- Thread starter bp2008
- Start date
Larebear
Pulling my weight
I played around with frame rates and bit rates a bit today to try to get a feel for storage requirements for continuous recording. At 10 FPS, CBR @ 10240, and BI set to cut the bvr file at 4GB, BI was rolling the files every 66 min.
With the above settings, BI was showing a bit rate for the camera of ~ 1250 kB/s.
Changing the Bit Rate in the camera to CBR @ 4096 resulted in a bit rate for the camera in BI of ~ 550 kB/s, with very little in the way of additional compression artifacts.
Changing the Bit rate in the camera to VBR @ 10240, with Quality set to 6 (Best), resulted in a bit rate for the camera in BI of ~ 160 kB/s, with noticeable compression artifacts. The interesting thing though, was that when I did an extract (no re-encoding) from BI and looked at the video directly, the artifacts seemed less noticeable. This makes me wonder if the BI console doesn't do quite as good of a job of displaying a VBR stream as it does with a CBR stream.
I used virtualdub to copy a key (I) frame at each of the above bit rates from the video, and then used irfanview to crop out the same section of each frame, showing a pitchback net sitting in front of a cinder block wall, about 45' from the camera.
VBR @ 10240 (actual observed bit rate of ~ 160 kB/s)
CBR @ 4096 (actual observed bit rate of ~ 550 kB/s)
CBR @ 10240 (actual observed bit rate of ~ 1250 kB/s)
VBR looks like it would really cut down on the file size, but I feel it is a little too aggressive in its compression, even at its best quality setting.
With the above settings, BI was showing a bit rate for the camera of ~ 1250 kB/s.
Changing the Bit Rate in the camera to CBR @ 4096 resulted in a bit rate for the camera in BI of ~ 550 kB/s, with very little in the way of additional compression artifacts.
Changing the Bit rate in the camera to VBR @ 10240, with Quality set to 6 (Best), resulted in a bit rate for the camera in BI of ~ 160 kB/s, with noticeable compression artifacts. The interesting thing though, was that when I did an extract (no re-encoding) from BI and looked at the video directly, the artifacts seemed less noticeable. This makes me wonder if the BI console doesn't do quite as good of a job of displaying a VBR stream as it does with a CBR stream.
I used virtualdub to copy a key (I) frame at each of the above bit rates from the video, and then used irfanview to crop out the same section of each frame, showing a pitchback net sitting in front of a cinder block wall, about 45' from the camera.
VBR @ 10240 (actual observed bit rate of ~ 160 kB/s)
CBR @ 4096 (actual observed bit rate of ~ 550 kB/s)
CBR @ 10240 (actual observed bit rate of ~ 1250 kB/s)
VBR looks like it would really cut down on the file size, but I feel it is a little too aggressive in its compression, even at its best quality setting.
Attachments
I moved the camera to a new temporary location. Here are a couple of snapshots, first one with WDR off and second with WDR on and set to the default 50. Finally got some sunshine today. Will post another capture later tonight. Camera has 2.8mm lens.
WDR Off
WDR On (50)
Just by saving your WDR shot and using a photo editor, I was able to increase the Gamma value, Contrast and Saturation making the image look much more natural....but the shadows are still almost non-existent and I would expect to see well-defined shadows even with WDR on....Since there's some puffy cumulus clouds in your shot, do you happen to know if by chance the sun was behind one of the clouds during the WDR shot? This looks more like the kind of image I'd expect in the shade.
Well spotted @wxman. I agree with you. I also bet that the sun was behind the clouds. The hard shadow in the middle of the first picture which is coming from the houses at the right of the picture does not exist in the second picture. Therefore the WDR feature feature of the camera has no shadows to compensate, in other words the contrast in the second picture is flatter than the first. However a proper WDR sensor would automatically compensate for the lack of high contrast areas and would not flatten (or reduce) the picture contrast. In other words it seems that the camera is trying to immitate the wide dynamic range by reducing the contrast all over the picture. I still have not seen any creditable demostration of the WDR ability of the new 4Mpixel cameras.
We also must understand that a true wide dynamic range sensor should not require any firmware alteration or software manipulation of the image.
We also must understand that a true wide dynamic range sensor should not require any firmware alteration or software manipulation of the image.
Last edited by a moderator:
Larebear
Pulling my weight
It was sunny when I took both snapshots, but it's possible that there might have been a cloud moving by on the second shot that partially blocked the sun. I'll try again on a clear day with no clouds. Today, it's all clouds and rain.
Next time I'll try tweaking the brightness, contrast, sharpness, etc. with WDR turned on, although I'm pretty sure I won't be using it here.
This is an unfair comparison. The HIK shot looks more blured than the dahua because the resolution of the HIK is only 2Mpixels rather than the 4Mpixels of the dahua. The dahua is certainly crispier due to the resolution.
Since we all know that the FOV in camera changes between different resoutions, at least configure the two cameras at equal resolutions.
Thanks! That would be very helpful just to make absolutely sure we're comparing apples to oranges. I know on my Hikvisions (3mp non-WDR) that the image can look perfect while the sun is shining but as soon as a cloud rolls over partially blocking the sun, my yard turns so dark that you almost can't even recognize the plants.
Speaking of clouds, one thing I did notice is that in the non-WDR image, the clouds are very over-exposed, whereas in the WDR shot, the clouds are normally exposed making the layered structure of them more identifiable. In my Hikvision Non-WDR cam, clouds get more over-exposed when a cloud partially blocks the sun because the darkness of the shady ground puts more range into the picture. When the sun comes back out in full strength, the cloud exposure is more normal....So if it's true that your WDR shot was while the sun was partially blocked, then that would still indicate some improvement as it was able to keep the clouds from overexposing despite increased range in the image.
Since I stated that it was a comparison of the FOV between two of my cameras at specific settings, I don't know how you can say it was unfair. Do the two pictures not demonstrate the FOV seen by each camera?
And by the way, on the Dahua, your statement that the FOV changes between resolutions would be incorrect. Unlike the Hik, which gives a different FOV for 3mp than it does for 1080p, the Dahua appears to scale the image to whatever resolution you choose. A snapshot taken from the camera gui at each of the 6 resolutions that the Dahua supports, all show exactly the same FOV.
Obviously it is a matter of priorities and compromises between resolution and FOV. I would select as priority the highest possible video and picture quality and compromise on the FOV. On the other hand I did not know that Dahuas scale up the image to the configured resolution, HIKs do not scale up. However does n't scale up burden the camera CPU since the scale up is a CPU intensive operation? Also does the scale up create any artefacts like jagged diagonal lines of high contrast?
Honestly I was a bit surprised when I looked at the snapshots from the different camera resolutions. The 1.3M (1280*960) looked squished. You can get a feel for what the images would look like if you take the 2560x1440 image I posted above and past it into irfanview, and then tell irfanview to Image>Resize/resample, clear the "Preserve aspect ratio" checkbox, and set the dimensions to each of the following pixel settings:
2304x1296
1920x1080
1280x1024
1280x960
1280x720
2304x1296
1920x1080
1280x1024
1280x960
1280x720
Larebear
Pulling my weight
Hmm, very strange how this WDR works. Sure seems like it just reduces the contrast of the image. I downloaded the images and put in a photo editor and can digitally enhance both images to look basically identical. I can't really distinguish much (if any) "range improvement" in the WDR. I must say that I'm disappointed and really wonder if this is "true" 120db WDR like it's supposed to be or just some type of digital WDR.
As a comparison, this is a Sony Effio-P Standard Def. Analog camera with 120db True WDR. Notice how with WDR off, it's so dark in the room that you can't identify the dude while everything out the window is overexposed/blown out. When WDR turns on, it instantly lightens up inside the room while darkening outside the window. This is the kind of results I was hoping/expecting.
As a comparison, this is a Sony Effio-P Standard Def. Analog camera with 120db True WDR. Notice how with WDR off, it's so dark in the room that you can't identify the dude while everything out the window is overexposed/blown out. When WDR turns on, it instantly lightens up inside the room while darkening outside the window. This is the kind of results I was hoping/expecting.
Larebear
Pulling my weight
Thanks for the comparison wxman. I have no experience with WDR and didn't know what the advantage was to have it enabled. As you mentioned, it seems like the WDR with the Dahua didn't do much more than decrease the contrast. Must be digital and not be true WDR right?
@wxman Well done and thank you for properly demonstrating the true WDR effect. You have confirmed my suspicions as well that this dahua may not really be true WDR. Untill someone produces similar to yours results from the Dahua, the Dahua should not be considered true WDR at the sensor level.
Last edited by a moderator:
Interesting, this really helps understanding what WDR does as the previous pics were confusing. Wonder if the Hikvision FWD models have true WDR or digital.
You're very welcome! I guess it's possible that true WDR can be accomplished in different ways in which some sensors may do it better than others. I'm assuming Dahua may use Aptina brand sensors, which may perform WDR differently than Sony brand sensors. Also worth noting is that with the Sony example, they convert WDR mode to "full" and contrast automatically changes to "high"...Perhaps similar results could be had with the Dahua if one were to manually increase the contrast when in WDR mode? One other factor to consider is that Sony Effio-P sensor is a CCD sensor, whereas the Dahua 4mp model is a CMOS sensor. Perhaps WDR works better on CCD than CMOS? More questions than answers here, so we can't say for sure yet that Dahua does not have true WDR; but from my initial impression, it isn't living up to my expectations.