H.264 High vs Main vs Baseline TestedBy: Derek Ward, Published on Jun 27, 2014
In surveillance, H.264 is often considered a single 'thing' but there are many different sets of capabilities, called profiles, to choose from.
The most well known of these are:
This chart shows the theoretical tradeoffs amongst them:
Basically, as you move 'up' from baseline to main to high, the theory is that you get the same quality for less bandwidth. Some vendors even claim that high profile reduces bandwidth by up to 50% compared to baseline profile.
In the beginning of H.264 IP cameras, baseline profile was the only option. However, over the past few years, more and more manufacturers have added in main and even high profile support.
In this report, we share our test findings comparing the bandwidth consumption of 4 IP cameras that support all 3 of these profiles to see what the savings, if any, really are:
- Dahua IPC-HF3101
- Hikvision DC-2CD864FWD-E
- Samsung SNB-5004
- Sony SNC-VB600
- Manufacturer implementation varied significantly, with some cameras not decreasing at all from Baseline to Main but dropping significantly when changed to High profile, while others actually increased when moving up profiles.
- Differences between main and high profile in low motion scenes varied widely from a decrease of about 50% to an increase of about 30%.
- In a high motion low light scene, switching from main to high profile resulted in differences which varied from reductions of 25% to increases of 25%.
- Differences when switching from baseline to main profile were typically small, less than 10 percent.
- CPU usage was roughly similar regardless of which H.264 profile was used.
- Observed latency was roughly the same regardless of profile.
- Selection of high, main, and baseline profiles varies between cameras, often with different names used to describe the same setting.
- Dahua IPC-HF3101: 2.420
- Hikvision DC-2CD864FWD-E: 5.0.2
- Samsung SNB-5004: 1.13_131218
- Sony SNC-VB600: 1.12.0
Here are our key findings from this test:
While common knowledge states that bandwidth decreases across the board as H.264 moves from baseline to main to high profile, unfortunately, this is simply not true. Users seeking to reduce bandwidth should test their camera in place with multiple profiles to determine which is best for their application, as manufacturer implementation and savings vary widely. Switching to high profile in some cameras may increase bitrate, i.e., make things worse.
Switching H.264 profiles is not supported by all manufacturers, with some defaulting to only one profile (typically baseline or main). Others support high profile in upper tier cameras, but not others. For example, Axis only supports high profile in cameras using their ARTPEC-5 chip, while others support only baseline and main.
Finding out exactly what profile is used may be tricky. A few manufacturers list it in the camera user manual or spec sheets, such as this example from the Axis P series domes:
However, if it is not listed, a stream analyzer such as AVInaptic must be used to determine which profile the camera is using. This is shown under the "Video bitstream" section in AVInaptic, seen below:
CPU Usage and Latency
As we found in our previous 2012 test of baseline vs. main profile, there was no increase in VMS client or VLC media player CPU usage when decoding high profile streams vs. main vs. baseline. Note that our test PCs all use dedicated graphics cards, and client PCs without may see some increase, but this will vary.
Visible latency between profiles was also unchanged. Some cameras showed more latency than others (typically well under a second), but this remained approximately the same regardless of what profile was selected.
Full Light Comparison
For this test, we used an interior full light scene, ~160 lux, with no moving objects, seen in the FOV example below.
We measured bandwidth with all cameras standardized to quantization 27/28, at 10 FPS. In this scene, Dahua and Hikvision both decreased when stepping up in profile, though Dahua saw the biggest difference from main to high profile, while Hikvision decreased only when moving from baseline to main. Both Samsung and Sony ultimately increased slightly when moving from baseline through high.
Dark with Lasers
Next, we tested the cameras in an extreme motion scene (<1 lux, with a DJ laser one) to see what savings, if any, were caused by changing profiles.
Again standardized to 10 FPS and Q27/28, Dahua and Hikvision decreased, while Samsung saw little change. However, Sony increased about 4 Mb/s between baseline and high profile, about a 20% increase.
Setting Profiles for Different Cameras
Finally, we break down a few examples of how to change between high, main, and baseline H.264 profiles for various IP cameras which support this feature. Below is a compilation of each cameras H.264 profile settings with their respective UI.
In Dahua cameras, profile is selected under "Encode Mode" seen below. H.264 without notation uses main profile (which is what the camera defaults to) while H.264B and H.264H use Baseline and High profiles, respectively.
Hikvision clearly and specifically allows the user to choose profile in its own dropdown, labeled simply "Profile."
Samsung also clearly labels profile selections, simply BaseLine, Main, and High, under Advanced encoder settings.
Finally, Sony does the same, with a dedicated dropdown for Profile in the Video Codec menu.
Camera firmware versions are listed below: