What is HDR?
HDR has appeared everywhere, we have:
- HDR image sensors,
- Multi-exposure HDR on image sensors,
- HDR processing within dashcams,
- WDR processing within dashcams,
- HDR image files,
- HDR display devices.
HDR simply means
High
Dynamic
Range. The dynamic range is the distance in brightness from the darkest to the brightest that can be recorded, and a high dynamic range is one that is larger than "normal", or these days it is probably more accurate to say that the range is greater than can be correctly stored in an
SDR (
Standard
Dynamic
Range) video file and displayed on an SDR TV.
6. HDR display devices
Starting with the display devices, because the display manufacturers give us a good definition of what HDR is, at least for display devices. The maximum brightness of a display device is measured in "nits", with a standard CRT TV (from last century) measuring 100 nits, and a modern good quality HDR TV rated at around 1000 nits, so 10 times brighter. In between, many standard LCD screens measure at around 200 nits, with many HDR LCD screens between 400 and 1000. Screens intended to be watchable in shaded sunshine, such as an iPhone screen, should be around 1000 nits, and outdoor TVs for use in stadiums etc. where they need to be watchable with the sun shining directly onto them are around 10,000 nits.
What is a nit?
A nit is a unit of brightness for displays, equal to 1 cd/m2, where a cd (Candela) is the scientific unit of measurement for "luminous intensity". Essentially a 1 nit display has a brightness of about 1 standard candle power. Moonlight also averages around 1 nit, although it depends on the phase of the moon. The sun is around 16,000 nits.
5. HDR video files
If you want to take advantage of your HDR TV then you need to play an HDR video file, because if you play an old fashioned SDR video on an HDR display device then it will display it with a 0-100 nit dynamic range instead of 0-1000. Presumably this is to force you into upgrading your video collection, your video player, your cameras, your phone, your computer, etc. etc. ?
There are several different formats of HDR video file, each with a different dynamic range, but the most common is known as HDR-10 and has a 1000 nit range. HDR video files use the same video codecs as standard files and have the same file extensions, so it is hard to tell the difference without inspecting the media info. Of course, if the file contains anything at 1000 nit brightness then it is going to be obvious on playback! The "10" in HDR-10 indicates a bit depth of 10 bits for the luminosity data, which is gives 4x the resolution of a normal 8 bit SDR video file, although covering 10x the dynamic range.
The A139 Pro does create HDR videos since it captures a very high dynamic range, but it stores them compressed into SDR H264 .MP4 video files. Hopefully we will soon have dashcams that store the video in 10-bit HDR video files, but for evidence purposes that is not actually necessary, and if you want to create nice HDR movies, you can use your video editor to convert SDR to HDR. The result of using SDR files is that the dynamic range gets compressed, contrast is reduced, and the video does not look as natural as it would if stored in an HDR video file and displayed on an HDR TV, but for evidence purposes, 8-bit detail compressed into an SDR file is sufficient, and SDR files have the big advantage that everyone can view them.
1. HDR Image Sensors.
All image sensors have different dynamic ranges, generally the larger the pixels the larger the dynamic range, so for dashcams, where we want a large dynamic range, big sensors are good. In addition Starvis-1 brought a decent increase in dynamic range of the sensor, and Starvis 2 has brought another 2.5x increase.
The Sony IMX678 Starvis-2 1/1.8" image sensor used in the A139 Pro has 2.5x the dynamic range of the standard A139 without using any multi-exposure tricks, so it can be considered a High Dynamic Range sensor, and it can also do multi-exposure HDR to extend the dynamic range further.
2. Multi-exposure HDR on image sensors,
The A139 Pro uses 2 exposures when the "HDR" setup option is enabled, so that could, at most, be an extra 2x the dynamic range, I assume there is actually some overlap.
The multi-exposure part of HDR not only extends the dynamic range, but it also allows for the brighter parts of the image to be captured with less motion blur than the darker parts, which for a dashcam is a big advantage. HDR is not only about multi-exposure images though, multi-exposure is not actually required to achieve a High Dynamic Range, but it does improve the results.
3. HDR processing within dashcams,
Currently, multi-exposure HDR requires that the exposures from the image sensor are combined into a single frame by the processor, or SoC (System on Chip), before being stored in the video file. Getting good results from the merging of exposures is complex, and on a dashcam has to be done in real time since the video file has to be stored onto memory card within three seconds. The only really good results I've seen come from the latest Novatek processors. They are still working on it, so there are likely some dashcam specific improvements to come. The aim for a dashcam is rather different to other cameras, since we are trying to freeze motion as much as possible, each individual frame wants to be as sharp as possible to record evidence, while other cameras generally want to keep some motion blur in order to smooth the transition between frames and make the video appear smooth when watched, so we need processing that is specialized for dashcams, which most SoC manufacturers are not interested in. Most video cameras set their exposure and processing based on the "180 degree rule" for shutter speed, but this is irrelevant for dashcams, and the resulting motion blur would severely damage the quality of the evidence within individual frames, even though the video might be nicer to watch as long as you play it back at normal speed without pausing to study the evidence.
4. WDR processing within dashcams,
WDR is a feature found on many dashcams, and is probably implemented in all decent dashcams even if they don't mention it as a feature in the specifications. Normally it just means recording the video with an increased gamma over standard, which means that the dark shadows have their brightness turned up, while the brighter parts of the image are left unchanged, this allows more evidence in the shadows to be squashed into the 8-bit SDR file. As the dynamic range of the sensor increases, more WDR can be used to pull detail out of the shadows without also pulling noise out of the shadows. As a result, the Starvis-2 sensors produce significantly better shadow detail than older sensors, although there is still a problem that shadow detail is often discarded by the compression codecs, probably because they were designed for older sensors where a lot of the shadow detail was noise and was thus unwanted. Hopefully we will soon see 10-bit codecs, along with 10-bit HDR files, and then we can use less WDR and still have the dark shadow detail available while having more natural looking videos.
