Why are dashcams video quality still so bad?

[Nigel]

No, higher bitrate means less compression hence less processing power which should mean less heat at the processor chip. The heat at the sensor itself is the same because the sensor captures a raw image

However it also means a lot more heat used to burn the extra data into the memory card and the camera ends up hotter.

 

[Kip]

Somewhere on the JooVuu X forums there are some test results using an X on standard firmware vs an X on modified firmware recording at a higher bitrate. When recording at a higher bitrate, the camera draws more current and produces more heat overall.

I don't know about CPU usage when recording at different bitrates, but, at playback time higher bitrates produces higher CPU usage on my laptop. It might be less compression to work through but that means more actual data to process, evident by the larger file sizes. At 30mbps I need to lock my CPU at full speed otherwise I get frame drops, whereas at 24mbps and 18mbps I can lock the CPU at a lower frequency and not suffer any frame drops.

 

[jokiin]

No, higher bitrate means less compression hence less processing power which should mean less heat at the processor chip. The heat at the sensor itself is the same because the sensor captures a raw image

low bitrate files aren't low bitrate because there's more compression, they're low bitrate due to less encoding and a lot of information being discarded, higher bitrate means more detailed encoding, more accuracy and more heat

 

[c4rc4m]

low bitrate files aren't low bitrate because there's more compression, they're low bitrate due to less encoding and a lot of information being discarded, higher bitrate means more detailed encoding, more accuracy and more heat

That is somewhat confused and against my understanding!

Jack is correct, uncompressed files / higher bit rate files need less processing as there's less to alter before they can be sent to be written. Try writing the same file in a video encoding program on your pc as an uncompressed avi and then as an MP4. You'll find the avi file is encoded and written in a fraction of the time of the same file written in Mp4 (or any other compressed format).

Low bit rate files are low bit rate because there's more encoding because the processor has to go through and examine, find and encode areas it can get rid of. It doesn't just get rid of them though, - if it did there would be blank areas in a picture. It codes them with a code (hence encoding) that tells the playback device what to produce in their place when they're uncompressed - so eg it may replace a cluster of pixels with a single colour found in several adjacent pixels (that's also why you can sometimes see "blocks" of pixels on playback in heavily encoded video - you're seeing very visible areas of pixels that have been replaced by code and then reproduced as single tone areas of replacement pixels when uncompressed. By replacing pixels with code representing those pixels, less pixel information is stored. It's kind of like the CPU being a secretary doing shorthand. More processing is required to create the shorthand but there's less to write.

The flip side is Nigel is correct that with a high data rate / uncompressed file, writing more data to the card will cause more heat as there's more to write. So on the heat side of things, what you save in one area you're going to create in another. Whichever area encoding or writing creates more heat will dictate which type of encoding - high or low bit rate, produces the most camera heat overall. From what I know, I'm guessing the writing side of things creates more heat as higher bit rates do seem to lead to more heat overall.

As for playback using more CPU, of course it will. With an uncompressed file, there's more information to process on playback if it's written in full instead of shorthand. Bit rate is an expression of the number of bits of information per second. With a lower bit rate there's less information to process and where encoded, the playback device produces areas of the picture from the information from the code instead of having to read information for every pixel (it can't of course because that information is discarded and replaced by the code representing it). With a higher bit rate file, there's more information to read and process on playback as it has to read the information for every single pixel for an uncompressed file, or for many more pixels for a high bit rate compressed file when compared to a low bit rate file. Again think of the secretary. If she wrote shorthand (low bit rate) there might only be 1 page to read. If she wrote it full hand, then the same information might result in several pages to read.

Again I stress I'm only an amateur, but this is my understanding.

 

[sludgeguts]

low bitrate files aren't low bitrate because there's more compression, they're low bitrate due to less encoding and a lot of information being discarded, higher bitrate means more detailed encoding, more accuracy and more heat

That is somewhat confused and against my understanding!

Again I stress I'm only an amateur, but this is my understanding.

I, honestly, don't have a clue about such things. I just shoved all my Mobius' settings onto max, inserted a 128gig card & let things run.
However, if I were in the market for such info, I'd be leaning towards that given by Jokiin.

 

[Bungus]

I, honestly, don't have a clue about such things. I just shoved all my Mobius' settings onto max, inserted a 128gig card & let things run.
However, if I were in the market for such info, I'd be leaning towards that given by Jokiin.

+1.
Years here now and still learning.

 

[Deleted member 46308]

+1.
Years here now and still learning.

Likewise

 

[c4rc4m]

I, honestly, don't have a clue about such things. I just shoved all my Mobius' settings onto max, inserted a 128gig card & let things run.
However, if I were in the market for such info, I'd be leaning towards that given by Jokiin.

..and me. Never knew about pixel binning until I read posts on this forum. Always informative.

 

[atarijedi]

I believe it is a 3 prong issue.

#1 is the frame encoding method, is it intra-frame (All-I) or inter-frame (IPB). Intra-frame is where each frame of the video is independent of the others, obviously high quality, implicitly requires higher bit rate, Inter-frame is where an Intra-frame is taken, then the next 2-5 frames are built based on changes to that initial Intra-frame, this is lower bit-rate, allows for higher compression. Most (if not all) dash cams, and even action cams, are inter-frame.

#2 is the bit rate, less bits is less data. Frame rate will also play into this, a camera shooting 1080p60 at 20Mbps will have a worse image than one at 1080p30 20Mbps because there is more data per frame at 30fps, not hugely as the encoding schemes are typically variable bit rate. The 30fps will have more motion blur though as well, obviously.

#3 is the field of view of the lens. In combination with the encoding method and the bit rate, the wider the FOV, the more image you are trying to squeeze into the frame, so data will be lost as encoding tries to keep the file size under control. Particularly small detailed items with similar colours that are close together, like the characters on a license plate.

I don't understand why a company doesn't make a high end camera that is 1080p60, 1440p60, or UHDp60, with a 50Mbps or higher bit rate, All-I frame encoding, and selling it for like $350-$400.

 

[Nigel]

I believe it is a 3 prong issue.

#1 is the frame encoding method, is it intra-frame (All-I) or inter-frame (IPB). Intra-frame is where each frame of the video is independent of the others, obviously high quality, implicitly requires higher bit rate, Inter-frame is where an Intra-frame is taken, then the next 2-5 frames are built based on changes to that initial Intra-frame, this is lower bit-rate, allows for higher compression. Most (if not all) dash cams, and even action cams, are inter-frame.

Given that the normal limitation for a dashcam is bitrate, using intermediate frames significantly increases the image quality for a given bitrate since it provides much better compression ratios. The penalty is the processing power required for calculating the change information, and since dashcams have to encode in real time that processing power is limited which in turn limits the image quality, however removing the intermediate frames would certainly reduce the image quality at the bitrate limitations we have.

#2 is the bit rate, less bits is less data. Frame rate will also play into this, a camera shooting 1080p60 at 20Mbps will have a worse image than one at 1080p30 20Mbps because there is more data per frame at 30fps, not hugely as the encoding schemes are typically variable bit rate.

More data per frame doesn't necessarily help with video quality when the extra frames are all intermediate frames. The image quality of the video as a whole depends on the data per second, not the data per frame. The extra frames just give for more temporal accuracy. Given the limited processing power available within dashcams you do need a little extra bitrate for 60fps v 30fps to give the same quality, but not much, and as processing power improves it gets less.

The 30fps will have more motion blur though as well, obviously.

The amount of motion blur is entirely dependent on the exposure time, the frame rate has nothing to do with motion blur other than allowing you to use a longer maximum exposure time in the dark. During daylight there is no difference in motion blur for 30fps v 60fps and with the latest high sensitivity sensors there is no difference at night under street lights either. If you see more blur in your 30fps video, it is probably because it is using a lower bitrate, less data per second means less detail.

#3 is the field of view of the lens. In combination with the encoding method and the bit rate, the wider the FOV, the more image you are trying to squeeze into the frame, so data will be lost as encoding tries to keep the file size under control. Particularly small detailed items with similar colours that are close together, like the characters on a license plate.

The amount of detail recorded is the same for wide and narrow fields of view, it is just that the wide field of view spreads the detail over a wider area and thus there is less within the area of your number plate, however the wide field of view can see the number plate when it gets closer without it falling off the edge of the image and then give you the same amount of detail as the narrow field of view did while it was further away. So wide and narrow fields of view is not about the amount of detail captured or image quality, but about how far away you want to look, how much zoom you are using.

I don't understand why a company doesn't make a high end camera that is 1080p60, 1440p60, or UHDp60, with a 50Mbps or higher bit rate, All-I frame encoding, and selling it for like $350-$400.

Because until recently that has not been possible, in fact Gopro still don't have a UHDp60 action camera yet, and since very few people would buy a dashcam at that price there is no reason to rush to get one out. However the latest reasonably priced chips do support 50Mbps and UHDp60 so we will see them arrive in the next year or two. However these specs do use more power and for a dashcam that creates a heat problem, the same problem you get when you use a UHD action cam as a dashcam and it doesn't shut itself down when it starts to get a little warm like a Gopro does:

 

[c4rc4m]

I believe it is a 3 prong issue.

#1 is the frame encoding method, is it intra-frame (All-I) or inter-frame (IPB). Intra-frame is where each frame of the video is independent of the others, obviously high quality, implicitly requires higher bit rate, Inter-frame is where an Intra-frame is taken, then the next 2-5 frames are built based on changes to that initial Intra-frame, this is lower bit-rate, allows for higher compression. Most (if not all) dash cams, and even action cams, are inter-frame.

Progressive is the answer, where each frame is a complete picture in it's own right. That's the only way you'll ever get a perfect still as it's not a mixture of 2 frames unlike interlaced. Full frames though contain twice as many lines and thus data as interlaced though.

#2 is the bit rate, less bits is less data. Frame rate will also play into this, a camera shooting 1080p60 at 20Mbps will have a worse image than one at 1080p30 20Mbps because there is more data per frame at 30fps, not hugely as the encoding schemes are typically variable bit rate. The 30fps will have more motion blur though as well, obviously.

True because with one you're trying to pack twice as many frames into the same amount of compression. However, many cameras up the bit rate for 60fps modes helping to alleviate this. Also, it's a matter of compression losses vs speed blur.

#3 is the field of view of the lens. In combination with the encoding method and the bit rate, the wider the FOV, the more image you are trying to squeeze into the frame, so data will be lost as encoding tries to keep the file size under control. Particularly small detailed items with similar colours that are close together, like the characters on a license plate.

True it doesn't help from that perspective. But the real issue with wide lenses is that non professional lenses tend to be non linear across the face of the lens, so you get increasing amounts of distortion and thus loss of sharpness at the edges. The answer here is a low distortion lens. However, considering the low distortion lenses cost around $100-120 retail on a M10 screw (based on the Ribcage sites lens selection), that's probably $50-55 trade and so it adds to the cost of the camera. I do believe it's possible though and that at least one manufacturer is looking into this following suggestions I made about putting an M10 socket onto the front of dashcams to accommodate this. It's important to understand though, that any resulting cam is unlikely to be low end and $50-100 to buy. Anyone who's ever put a quality lens on a cheap SLR camera, will know just what a difference a quality lens can make. Conversely, you can buy the highest end camera body, but stick a consumer lens on it, and you won't get magazine quality shots. The lens is one of the key components. However, goods glass doesn't come cheap!

I don't understand why a company doesn't make a high end camera that is 1080p60, 1440p60, or UHDp60, with a 50Mbps or higher bit rate, All-I frame encoding, and selling it for like $350-$400.

If you take a look in some of the sub forums on here, then you'll find there is experimental 3rd party firmware with rates in excess of 50mbs. Some of the action cameras also have data rates of 100mbs. However, pure data rate alone isn't the answer. The issues are a combination of data rate, lens, sensor, processor, metering and cost in my opinion (might not be a comprehensive list as quickly put together and I am an amateur not a camera expert).

Also, so far as cost goes, I doubt you'd sell many Dashcams at that level. Personally I see the top range being $200-250 or £200-250. That would be top end and so probably not huge selling but I would imagine still viable as it's not beyond the top black box style dashcams now. Beyond that in price, I think anything you'd get would be very niche. Personally, I'd trade some features for image quality.

 

[HeWhoMustNotBeNamed]

Progressive is the answer, where each frame is a complete picture in it's own right. That's the only way you'll ever get a perfect still as it's not a mixture of 2 frames unlike interlaced. Full frames though contain twice as many lines and thus data as interlaced though.

atarijedi is talking about encode settings / methods.

it has nothing to do with interlacing. since DVDs came out nothing is interlaced anymore.

 

[c4rc4m]

atarijedi is talking about encode settings / methods.

it has nothing to do with interlacing. since DVDs came out nothing is interlaced anymore.

My Canon video camera offers a choice of encoding (recording) video into 25P or 50i formats onto it's media (it's no more than a 5 years old).

Also, the H264 standard supports interlacing: https://en.wikipedia.org/wiki/H.264/MPEG-4_AVC#Features

Features

H.264/AVC/MPEG-4 Part 10 contains a number of new features that allow it to compress video much more efficiently than older standards and to provide more flexibility for application to a wide variety of network environments. In particular, some such key features include:

................

Flexible interlaced-scan video coding features, including:

• Macroblock-adaptive frame-field (MBAFF) coding, using a macroblock pair structure for pictures coded as frames, allowing 16×16 macroblocks in field mode (compared with MPEG-2, where field mode processing in a picture that is coded as a frame results in the processing of 16×8 half-macroblocks).
• Picture-adaptive frame-field coding (PAFF or PicAFF) allowing a freely selected mixture of pictures coded either as complete frames where both fields are combined together for encoding or as individual single fields.

So, so far as I'm aware, interlacing is still in use on some cameras.

Even with Intra Frame though, you're not going to get a perfect picture so far as I'm aware as the encoder only encodes obvious changes to the picture in each subsequent frame with reference to the original frame. Fine if the rest of the frame hasn't moved or changed in the slightest. However, again so far as I'm aware, small changes are allowed without being recorded (otherwise you'd be back to encoding the whole of every frame), and that results in most frames not being perfect stills. I stand to be corrected though.

As an aside, some interesting points raised from an article on CCTV here on how to achieve the best picture: https://www.ifsecglobal.com/cctv-camera-settings-heres-how-to-get-the-best-picture-possible/

 

[tabetha]

I thought the iphone5 did a good job for a phone, but it is light years behind my thinkware F770 and I suspect many other dashcams

 

[Moon]

Speaking of Video Quality .. Which Uploading site do you folks use to Upload your Dash Cam Vids to show in this Forum?

Is there one that stands out as better than the others .. and does not want to know all of your personal details like your contacts etc?

I want to post up some of my journeys and would like better Videos than YouTube can do.