The Worst HD Quality Camera Ever (Samsung 1280H)

Author: John Honovich, Published on Oct 15, 2014

Samsung, what are you doing?

Their new 1280H offering significantly stretches video and is one of the most bizarre offerings we have ever seen.

Earlier this year, Samsung trumpeted that 1280H would "allow HD quality video to be deployed without the need to rip and replace analog infrastructure.” We bought the 1280H SCB-5003 box camera and its companion SRD-476D DVR so we could test this out.

Here is the 'original' / 'regular' sized Samsung image from that camera / recorder (click on image for full size):

Good quality for analog, no problem.

Here it is in Samsung's special 1280H (1280 x 480) mode (click on image for full size):

Both are the exact same scene, same time, just different resolutions.

The 1280H resolution stretches the image by 81.8% horizontally, though Samsung denies this, see their response at the end.

Get Video Surveillance News In Your Inbox
Get Video Surveillance News In Your Inbox

[UPDATE: Samsung's counterargument appears to be that they do not stretch it 'from' 4CIF but that they distort it when they encode it from analog. So they take a ~4:3 scene captured and encode it at ~8:3 aspect ratio. Relative to SD analog, the image is stretched. Relative to 1.3MP, the image is squeezed. Either way, it is distorted, as shown above and below.]

To show the extreme stretching, this comparison focuses just on the test chart and the subject:

Notice how much wider the chart, license plate and person in the 1280H version clearly are.

Video Demonstration

The screencast below shows how Samsung displays normal and super stretched images and video:

Stretching Problems

It is obvious that Samsung 1280H distorts the image / video extremely wide. It warps the dimensions of objects and people, increasing the likelihood of evidentiary problems as the video does not accurately represent the scene. Furthermore, increasing pixel count / 'resolution' by stretching video is misleading.

Worse Than 960H

Those of you who have had the misfortune of using 960H will immediately notice it is the same absurd trick, simply taken to the next level by Samsung, stretching even wider.

Why Do This?

Theoretically, Samsung can now claim that they have increased the resolution of analog cameras from the common 4CIF 704 x 480 (.337 MP) to 1280 x 480 (.614 MP) and then try to claim 'HD quality'.

UPDATE: Better Than SD If Stretched Out

At 1280 x 480, as captured and displayed, the video is distorted. However, if you manually stretch out the 1280 x 480 image to the correct aspect ratio of 1280 x 960, it shows moderate quality improvements over a 4CIF version. However, even with 1280H stretched out, it shows noticeably less details than a true HD 720p camera. Comparison:

No Comparison To Real HD

We tested Samsung 1280H against other analog and real HD cameras (Samsung own WiseNet III models).

As the comparison below shows, 1280H is similar to slightly better than analog but notably worse than either HD camera:

Samsung Responds

Prior to publication, we shared our concerns and test images with Samsung.

Here is their response:

1) 1280H is not image stretching. Camera make more high frequency image with detail information. And DVR can up-sample the signal to keep image quality and detail information.

If you compare recorded data from 1280H and legacy D1 DVR, 1280H DVR shows better image quality.

2) Vertical line of 1280H product has 480 (576 for PAL) to keep legacy analog product. So, these product are still compatible already installed analog product without any gender or adding product.

3) When you playback in DVR directly or with Client S/W, it shows 4:3 video with stretch vertically.

Additionally, they provided 2 images: TVL Comparison Chart and a 1280H system architecture diagram [removed later at Samsung's request].

IPVM Final Thoughts

It is a shame that vendors are pushing stretched 960H and now super stretched 1280H analog.

We think the market will quickly see through this. The good news is that real analog HD options exist, like CVI (tested), AHD (tested) and TVI (up next for testing). Maybe Samsung can release one of those...

Key Findings

Here are our key findings from this test:

Pricing

Pricing on the 1280H DVR and camera used in this test are as follows:

This is notably more expensive than many high TVL count analog cameras, the most expensive of which was ~$60, and even more than HDCVI cameras, which typically sell below $50.

1280H Aspect Ratio

Stretched 8:3 nonstandard aspect ratio vs. SD 4:3

Zoomed on subject,difference even more obvious.

Vs. 960H

Again, zoomed in on the subject.

All three, side by side.

Performance vs. HD and High TVL Analog

In full light, not stretched, captured at 4CIF resolution, the 1280H camera provides video similar to 900TVL, better than 960H. 720p and 1080p obviously far, far better.

Low Light Performance

SCB-5003 low light performance similar to HD Wisenet III models (obviously lower resolution)

3.2MB .zip file of these snapshots

1 report cite this report:

Winners Losers Fall 2014 on Oct 13, 2014
We are all losers, IPVM included. The historically poor new releases for ASIS / Essen make 2014 the worst year for surveillance innovation in, at...

Comments (79)

Only IPVM PRO Members may comment. Login or Join.

So, how many typical end users are going to look at 1280H and think something is wrong with the camera?

The first salesperson who tries to tell a customer 'this is better' gets laughed out of the room.

Interesting.

Can you clarify your definition of "stretched"?

Does it mean stretched TVL lines or stretched pixels? Or does it mean they added an additional 100 (For 950H) TVL lines or pixels?

If it is adding lines/pixels, is it similar to the way a Blu-ray player upconverts a standard DVD? The Upconversion looks a lot better than the standard DVD but it not up to par with a HD movie even though it is getting closer with for the average human eye.

When using a 960H camera without a 960H DVR, it will be stretched as the extra 100 (950H) lines are not there...Correct? (How would it look better if the DVR only records the same amount of lines)? The main part I would like to know is if the 960H camera or 1280H camera with the 960H or 1280H DVR actually add TVL lines/pixels like they are supposed to or if they are just stretching the image. If it is just stretching the image, then it would be false adverting on the manufactures part.

When I work with Pixels in Adobe Photoshop and enlarge an image, it does 2 things. It adds pixels to each color and removes the added pixels on the edges that create soft colors between each color. Sometime you need to define the edges better by getting rid of the soft colors or making the soft colors into hard colors or do a little of both. It is also better to enlarge pixels 10-15% at a time as it helps with quality. I believe this is similar to how Blu-ray players upconvert without enlarging pixels 10-15% at a time.

It's about pixels, not lines.

You can do the same thing in Gimp, for example taking the regular Samsung image and scaling it up to 1240 x 480:

Yes, this image has more pixels (81.8% more) but it's worse than useless because to do this you are making each object in the scene wider, distorting what images look like in reality and adding no new details.

I just put both in Photoshop and noticed something interesting when I took the image from 100% to 1600% on the screen (Not increasing actual image size) You can still read the licence plate with the 1280H image and you cannot read the license plate on the 704x480. Below are the images snipped of my screen with the snipping tool. 1400% is the same way except the 1280H image is completely clear when reading the plates. What is the benefit? Don't know? Maybe to zoom in on something real small in the origianl image?

1280H (1600%)

704x480 (1600%)

And as a comparison, same scene, same time, here is the license plate excerpted from the 720p Samsung camera:

Far clearer and correct aspect ratio.

Also, interestingly, if you take the 4CIF original, scale it horizontally 81.8% in Gimp, you get the same basic plate picture as you do in Samsung 1280H mode:

Well, correct me if I´m wrong...

For me this looks like the old 720x240 images (2 cif?), where you have it widely stretched, but what you need to see is if this image is stretched from a 352x240 source, meaning pixel interpolation in width, and consequently no image quality gains because you are just making new pixels out of existing ones, or if the image source is actually captured as 720x240, meaning that the sensor has actually captured more pixels for each line of the image.

If the sensor actually captured more pixels for this 1280H and it is not just a marketing stunt (if the camera is actually strectching a 720x480 image into 1240x480) so I don´t see why it does not offer the benefits they announce, because when you view this camera in any VMS (or DVR?) software, you will probably not view it as a stretched image, but the VMS will correct the proportion (Usually filling the whole camera tile on screen) which will give the proper image (4:3, 16:9...), so, basically the software will have to stretch it up, but if you actually have more pixels in each line of the image I can clearly see the benefits.

And viewing the video, where you show that the samsung software will show it as original image (rescaled in proportion to original resolution), my guess is that the software must have some option to rescale the images to fit the tile instead of keeping the proportion to original resolution.

So, in my point of view, this is just like the old 2CIF resolution but bigger

"If the sensor actually captured more pixels for this 1280H"

The sensor does capture more pixels. They list a 1.3MP sensor.

The problem is it is transmitted by NTSC / PAL, which is the key limitation.

They could use a 10MP imager but the bottleneck is the transmission.

Eric, At least in theory, 720x240 is not stretched or squeezed. By definition, it is field-only recording/display. An analog video signal is typically interlaced, which means each full image consists of 240 odd horizontal lines followed by 240 even lines to complete the image. The scanning process leaves gaps between each odd line that are filled by the subsequent even line. 2CIF displays only the odd or the even lines, leaving the gaps.

"It's about pixels, not lines."

Are you totally posative about this? My understanding is that it is all about the lines (TVL) than the pixels. The analog camera collects the information with TV lines. The CCD image sensor takes the analog voltage and converts it to digital. With 960H and 1280H, it is adding more TVL's before it converts to a digital (Pixel) image hence why the finished product has more pixels.

In this case, the video / image is scaled up from 704 x 480 to 1240 x 480.

wow. thats just wrong of them. rather insulting when you think of it.

The 1280H image has a resolution of 120ppi. The 4CIF image has a resolution of 72ppi.

It is weird how the 1280H image has a higher resolution. Usually this happens because the image was Resampled (Adding pixels). However resampling usually keeps the same dimensions.

Figured it Out (Not scaled or stretched but resized and resampled)

I just figured out how they did it. They did resample it. However, they needed to keep the height at 480 (Legacy Product).

So to reproduce the 4CIF picture, increase the image size by increasing the resolution to 120 and then resizing the height to 873 (Width shoud go automatically to 1280) and resample by using a Bicubic Smoother. Either the resolution is created by increasing to 120 or it is from the increased pixels from the TVL. However, by increasing the height to 873, the picture is balanced. (873 pixels would be the equal height dimension)

"However, by increasing the height to 873, the picture is balanced. (873 pixels would be the equal height dimension)"

Agreed. And this is something Samsung mentioned potentially doing in a future release / upgrade.

But the question remains - what is the worth of taking lower resolution images and scaling them up?

You can increase the height / scale up any image. Take a VGA image, put in to a graphics editor scale it to 5MP. You will get an image with 5 million pixels, but that does not mean it is equivalent to a true 5MP camera.

This is megapixel alchemy....

I'm guessing that the fact you can use standard analog cameras with it is the appeal for the lower end market. If you listen to small security dealers, they will tell you that the customer needs/wants a system that is compatible with standard analog cameras. Something that SDI, CVI and TVI has not succeeded in doing unless you buy a hybrid DVR that is basically a computer for mucho dollars. Also - This market has waited forever for something that will work with their existing cameras. CVI and TVI rumors brought promise...but nothing for their existing cameras.

Manufacturers need to put out a product that fills a need. This need is way higher than I ever imagined. I used to think it was not even a niche. Now I know this market is a big niche.

Analog HD DVRs support analog cameras. The Hikvision TVI DVR we are testing does, as well (and IP cameras), but I'm not through testing it yet so can't comment on any issues or performance. With HDCVI you need a different recorder, true, but these are both low cost recorders (<$500) priced below Samsung's 1280H model which support analog, CVI, and IP.

To say that CVI and TVI rumors brought nothing isn't quite true. These HD over coax technologies are all very young (Hikvision has been shipping less than a month). It's far too early to tell how they'll perform with existing analog cameras.

Hi FWIW we have been testing and selling some OEM Hikvision TVI recorders and cameras since they became available a very short while ago. The lower end recorders, limited 1080p framerates(12fps across 16ch), are <$300 without hard drive, take 16 channels of analog or TVI in any combination and 2 IP. I currently have had one of these running for about 3 weeks with a 1080p Hik IP PTZ, one 3MP fixed dome Hik IP, 7 analogs varying and 7 1080P TVIs. Absolutely no problems with functionallity or performance. Just sharing, I think they hit the mark on this segment and need. Perfect? no but finally where lower end CCTV should have been for a while IMO.

The 1280H thing is a joke compared to these other technologies.

Well, I'm sold. This is why you ask your integrator to show you what this "new camera" looks like. I wish this article was visible to end users.

One, this article is public. Two, there's ~1,000 IPVM end user members as well.

I was not aware. Thank you for correcting as I believe this is useful to end users.

Sure thing. Btw, as a member you can tell what articles are public by checking right underneath the title, like so:

As a point of interest, I took the 2 images you posted and resized the 1280H image horizontally to 704 pixels, (removing the stretch distortion), in photoshop. If you then compare the 2 images, the 1280H image clearly resolves more horizontal detail. So I think its more than just a marketing trick.

B, thanks for the response. I upscaled both the 1280H and 4CIF versions to 1280 x 873, keeping / restoring the aspect ratios.

Here is what they look like:

The license plate is easier to read on the 1280H version. Which face is better is more debatable.

Of course, neither are close to a real HD camera.

John,

I agree that a 720p camera would be better than the 1280H camera.

However, I believe it is unfair to say that the image is "stretched" when it is not. Stretched is what happens when you actually stretch a physical object. Stretching an image is actually distortion.

At first glance, the image does look stretched. However if you study it, you will see it has more detail. In fact, if they did not have to worry about legacy product, they could make the image look normal by adding the appropriate image height (873). The word I would use would be “smooshed”, squashed, suppressed, scrunched or etc. (The problem is vertically and not horizontally) Like Joe C says on a Kid Rock song…(Vertically Challenged)

Below is an example of what happens if you stretch the 4 CIFF and not resample it using a bicubic smoother algorithm. At first glance it looks the same as the 1280H. However, if you study it, you will see a lot of missing detail. Even your 4CIF example, that used a bicubic interpolation algorithm, has more detail than the stretched version of the 4CIF.

Jeremiah, we agreed earlier that increasing the height would make the image look normal. But then you would have a 1280 x 873 image that would be fake megapixel. We solve one problem by creating another one.

You can certainly say its squashed, but that's relative to a fake megapixel version.

If Samsung turns around and says its now running 1.1MP (1280 x 873) over NTSC / PAL, then the situation is even more misrepresented.

It is definitely not MP. But Samsung does not market that it is MP, or have you seen different? From what I have seen, they market it as 1,000 TV Lines (High Resolution). Now the CMOS sensor may collect those pixels but when scanned...different story.

I would not buy this camera at this point, however, I would put the idea on the back burner. If it was another manufacturer who sells mostly by OEM'ing, I would throw the idea out the door. But this is Samsung. They may have something brewing in one of their Divisions that may play a crucial part in the future. If not, then I have no idea what they are doing.

They market it as 'HD quality'.

My point is, if they take your advice of increasing pixel count to eliminate the distortion, it would technically be a MP camera (1280 x 873 or 1290 x 960 or whatever pixel count they choose). My point is, at that point, it would become even more misleading.

Samsung has done well with WiseNet III as real IP cameras. Likewise, they should simply move to a real analog HD technology and eliminate this half baked attempt.

If this had been 1280x480 without any stretch, just massive extra width (you know like everyone thought 960H was) then perhaps they had something. But an evolution of 960H, oh dear. It's going to struggle against analog HD in terms of resolution - for openers.

Pricing will tell. If it's positioned slightly above 960H but way below analog HD then maybe they have something. But the CVI and TVI cameras are very cheap and give better results. So that's going to be difficult.

I guess it doesn't do the 300M+ distances of analog HD ?

Here is an article from 2012 talking about 960H which mentions 1280H.
It's interesting to look back at the technology from where we stand now and the direction all the players mentioned (inc Hik, Dahua, Samsung) have taken.

C, thanks for the feedback.

Pricing on the 1280H DVR and camera used in this test are as follows:

At those prices, as you mention, 'true' analog HD options will have both a video quality and cost advantage.

It reminds me of the consumers who buy wide-screen HDTVs and stretch SD pictures because "Dammit, I paid for a wide-screen TV and I want to see everything in wide-screen."

Good analogy with HDTVs, though this is even more extreme.

From SD to HDTV, it's a 'stretch' from 4:3 to 16:9 or ~33%.

From SD to 1280H, it's a 'stretch' from 4:3 to 24:9 or closer to 100%

http://en.wikipedia.org/wiki/21:9_aspect_ratio

Also at CES 2014, Samsung presented a 105-inch curved LCD TV with 5120×2160p as well, the other first UltraHD 21:9 screen.

Almost, but not quite :)

Yeah but that's display, not capture.


The 105" 4k display is real. The problem in broadcast World is that there is almost zero 4k content. I was at a Costco a month ago and they had one of those curved 4k displays playing a 4k rendering of Monsters University and it was just STUNNING to see. But you can't buy that 4k version of the movie, and the next-big-thing after Blu-ray is only in the research stages right now. There's only one company that makes a 4k broadcast-grade camera right now, and it's only sporadically being used.

It's important to note that in the Broadcast World, everything used to be done in film, so it was always possible to go back and re-master a digital recording of the film in 1080p, (and perhaps even again in 4K).

But for digital age movies that were captured by 1080p cameras, the only option is upsampling to 4k, which as most of us know is at best a band-aid. The real step-function won't occur in broadcast until content is actually being FILMED in 4k.

Movies that were digitally created (e.g. Pixar) can always go and re-render at a higher resolution. But live-action movies are limited to whatever they were originally captured at.

Now, none of that relates to security per se, but I was just addressing your point about displays.

You didn't address my point about displays. Admittedly it was a tenuous comment I made, but also a lot simplier.

At least Samsung have a display with an aspect ratio they can use with the 21:9 format of the 1280H solution.

4K (another topic elsewhere) agree with you entirely for consumer.

C,

The fact that Samsung makes a 21:9 display, doesn't mean it would display the video from a 1280H camera with the correct proportions. In fact, I calculate that the video would appear even more "stretched" if the image was adjusted to fit the display's aspect ratio. This is illustrated by how the picture appears on a 16:9 display.

Samsung's 21:9 displays appear to cost $1600 to $3000 so I am betting it's fairly immaterial to buyers of pseudo HD cameras aimed at the low end of the market.

Thanks for this analysis John. I was trying to think of a good analogy for this whole matter, but it's so absurd that I lost interest.

I'm going to guess more end users will fail to see through this than you might think, and that A&E's will put this 1280H in their specs without realizing that it's not what it claims to be.

History (e.g. low light specs, resolution specs, WDR specs, etc.) tells us that this is true. We are in an industry where spec trumps reality.

Jason, I can very well see it happening.

This is what I can gather, lookin at all the information, just my take.

  1. Dvr is stretching the video, VERTICALLY
  2. The reason it looks stretched horizontally is on account of the dvr hasn't fixed it by stretching it vertically.
  3. When y'all see it stretched that means "Your doing it wrong.", use the dvr client, not smartviewer.
  4. The reason it needs to be stretched vertically to begin with is because the PIXELS ARE NOT SQUARE on the sensor. They are taller by a good bit.
  5. Any software which treats the pixels with equal size will looked horizontally stretched.
  6. NTSC horizontal resolution is not hard bounded like vertical resolution, so it just may be possible to use squashed pixels to get more detail.
  7. The claim is that extra horizontal pixel information is modulated into the analog carrier and then extracted by dvr using oversampling
  8. Sony and Panasonic among other folk make these asymmetrical sensors, and make the same claims, (at least about 960H), so are we saying that all of them all is fibbin?

In short, to give Samsung a fair shake, we should just do a simple test in the DVR with one of them TVL charts and see what it reads, and then compare it to a top notch SD cam.

Since that is their claim why not take em up on it?

Jim, whether one says it is being stretched vertically, horizontally, squeezed, squished, etc., the bottom line, in EVERY case, is that the 1280h video is significantly distorted, relative to reality.

That is a critical quality problem and it is one they swept under the table, just like the 960H providers. I cannot imagine people who want 'higher' quality based on significantly distorting the objects in the scene.

As for Samsung's TVL chart test, it is obviously invalid just looking at the images provided, excerpted below:

Notice how both 1280H and 4CIF have the exact same perspective / aspect ratio. This is not possible on the recorder side, because 1280H is 1280 x 480 and 4CIF is 704 x 480.

Now, perhaps Samsung is claiming 1000TVL on the camera side before sending it out (NTSC / PAL) and encoding it on the DVR. But, in reality, what counts is what is recorded. And we have all of those comparison tests above that show this Samsung is similar to slightly better than analog cameras but significantly worse than true HD ones.

Notice how both 1280H and 4CIF have the exact same perspective / aspect ratio. This is not possible on the recorder side, because 1280H is 1280 x 480 and 4CIF is 704 x 480.

But, Sir, it is possible if

the sensors pixels ARE NOT SQUARE!

the sensors pixels ARE TALL RECTANGLES!

Then there is NO horizontal stretching, as such. Stretching meaning the scaling of a lower resolution to a higher one. When you see "stretching" that happens because the input pixels are being mapped to output pixels at a 1:1 aspect ratio. But only the DVR "knows" that source pixels are not square and can compensate for it (by stretching the vertical to the right aspect ratio.)

Jeremiah done hit the nail on the head:

At first glance, the image does look stretched. However if you study it, you will see it has more detail.

According to Sony, a 960H sensors pixel elements have an aspect ratio of ~10:15. Figuring that a 1280H is even more fangled, then it should make a little more sense.

And heading you off at the pass, let me interject also that NTSC's horizontal resolution is not constrained by 4CIFs 720/704 count, it is only limited by the signal bandwidth.

So on the theoretical account, it seems possible for some gain, but if we are convinced that its just stretching then we are not gonna look so hard right?

<edited above from 15:10 to 10:15 aspect ratio>

Jim, stop the theoretical speculation. Look at the actual test result output.

This Samsung camera / recorder offers two modes: 1280 x 480 and 704 x 480.

The dimensions of objects are different in the 1280 x 480 mode than the 704 x 480 in our test of their actual product.

Now, Samsung's TVL examples have the objects with the same dimensions. That is not possible with how they output / display video with their recorder.

Please stop with the semantics of whether its vertically stretched or horizontally stretched or squished or squeezed. It's clearly distorted, however you state it.

Are you seriously arguing that Samsung's 1280 x 480 mode does NOT distort objects as they appear in reality?

Beggin' your pardon Mr. Honovich, but you need to go back and watch the actual test result video yourself.

When Ethan is in the DVR client software (at 00:31) he clear as a bell says that the stream that "we're using is 1280x480, but we're not getting that stretched aspect ratio." He should be happy, but he is not because he thinks that he not getting the 'full resolution' because the video is not stretched. This is what anyone would think if they dont know the sensors pixels are extra narrow. He actually is getting the full resolution!

Then he snaps a screenshot and views it in an external viewer and says it is distorted. But as I explained twice already to y'all, what you are seeing is the viewer showing pixel for pixel the output of the sensor. And since the size of the pixels on the monitor are much wider than the sensor's pixels the whole image gets much wider when viewed that way. I reckon samsung could have expanded the vertical on the snapshot, and its easily done if desired. But anyway its just the plain ol bytes of the sensor in a file.

Back to the video: Not satisfied that he was getting the full resolution, Ethan then uses a DIFFERENT piece of software (1:30), samsung smartviewer, in search of wide/full resolution. He gets it. Now he is happy. Unfortunately as samsung has pointed out above, to gander the corrected aspect ratio you must use the special dvr software.

So maybe smartviewer is lagging behind with full support for 1280H dvrs. Noted.

As for the worse quality in the comparison, how EXACTLY was that shot made and why is the FOV a little wider than the 720p one? And its still only 480 v res, so in that direction its gonna be worse than HD, and maybe it doesn't quite make it up in the other direction but mistakenly accusing Samsung of just 'stretching' the video and insuiating they are lying about it is out of line, and as a matter of personal honor, should be retracted.

Jim, simple question which you've completely ignored:

In 1280H mode (1280 x 480) is the video distorted relative to reality? Yes or no.

Let's start with that. Try a one sentence answer and then we can discuss the rest.

No, not distorted, because as far as I'm aware the only software that has a '1280H mode' (meaning properly interpreting the 1280 x 480 datapoints which define the 1280H format) as such is the approved DVR software and there it is fine, as shown in the video.

Jim,

Here is how it is displayed in the approved DVR software. It is obviously distorted and not 'fine':

That is the same display / distortion that occurs when downloading the full resolution as a snapshot.

Samsung's distortion is consistent any time they show the full 1280 x 480 resolution.

John,

You are using American pixels, not Korean pixels. Therein lies the problem. American pixels tend to be obese due to over-eating and lack of exercise so they are square. Samsung's pixels consume less energy and exercise regularly so they are much thinner.

So, like the Seoul subway, Samsung can pack more pixels in the same screen (car) width than say New York (see below):

Seoul Subway New York subway

When you playback in DVR directly or with Client S/W*, it shows 4:3 video with stretch vertically.

The DVR directly or [Web] Client S/W are the two methods approved in Samsung's statement. Is that there image taken from Smartviewer? Ask Samsung when they all plan to support native 1280H in Smartviewer.

By golly, have a look see at the sensor of the camera yourself, is the horizontal imager size almost 3 times as wide as on the vertical? I wager not.

If it helps, then maybe think of the dvr software like dewarping, only instead of correcting for a lens asymmetry, its correcting for a sensor asymmetry. Not any old client can dewarp any fish-eye right? So Smartviewer doesn't have the 'dewarping' software built-in yet. BFD.

Outside of Samsung's customized 1280H viewer (like picture manager), the images will need PAR correction, PAR=Pixel Aspect Ratio. Don't confuse display ratio and pixel ratio. This type of adjustment is common in mainstream videography. Like Jeremiah's approach, but even more direct, take the snapshot from the web client into photoshop, create a custom PAR of .55 (under View) and you will get the correct aspect ratio. See that photoshop still shows the image as 1280x480. Change up the view to 100%. So there is the FULL resolution at a 4:3 ratio.

*I believe that the gentleman from Samsung is refering to the Web client software for the DVR, if you believe he means any client software then please ask him.

"is the horizontal imager size almost 3 times as wide as on the vertical? I wager not."

And that's exactly the point.

This has nothing to do with the size of the pixels.

It has everything to do with FoV aspect ratio vs the captured aspect ratio.

The imager aspect ratio is ~4:3 (spec'd at 1,312H x 1,069V). As such, the camera is capturing reality in a ~4:3 aspect ratio.

The 1280H resolution is ~8:3 (1280 x 480). Since the camera is encoding reality at the nearly twice the width of reality, the resulting image is stretched relative to reality.

Let's cut to the chase, do you still stand by this statement?

The problem is it is transmitted by NTSC / PAL, which is the key limitation.

Yes or No?

Of course. Why don't you think Samsung simply transmits / encodes record at a full 1.3MP? Why do you think they are creating this wacky 1280 x 480 stream? It is because of the NTSC / PAL limitation.

Jim, you'd do well to acknowledge the obvious - Samsung 1280 x480 is distorting the created image relative to reality. You can frame it is as squeezing or stretching or squishing, but 1280 x 480 is obviously and fundamentally distorted relative to the reality it is attempting to capture.

Why don't you think Samsung simply transmits / encodes record at a full 1.3MP?

Here sir is the answer, I dare you to read it and try to actually understand it before dismissing it.

NTSC's vertical resolution is hard limited to 480 lines. That on account of the electron gun sweeping back and forth, drawing 480 lines from top to bottom. That the NTSC convention. End of story for vertical res, its 480.

Horizontal resolution is a horse of another color. It has no fixed integer resolution limitation. Instead it is constrained by the bandwidth of the signal (~4 Mhz) and sensor pixel size, (or vidicon electron beam density back in the day). That there is one reason that hq b&w monitors can produce 1000tvl, (aided of course by the fact that they have no color grile).

Back to our case. See that 4:3, 1.3MP sensor? Its basically 1280x960 right? Problem is you can get all the horizontal resolution but none of the vertical. So what can you do? You take those 960 lines of 1280 pixels and you combine every two lines (by binning) to create just one line of 1280 pixels for the pair. (This line binning has the effect of having double height pixels). You may need to think a sec about this. Now you have a 1280x480 which CAN be sent using NTSC.

So then you modulate the 1280x480 and send it down the wire and everything is ok, right? Almost, but because of the 2:1 vertical binning which you did, any monitor not taking this into account ends up making it twice as wide as it should be. But by simply doubling every input line you will recreate the horiz. resolution and aspect ratio of the imager. And that is why you need a viewer that knows to do this. Its not distortion, its called encoding, and without the proper decoding it just don't look right, just like anything not decoded right. Sure it would be great to have the vertical res too, but it just not possible with NTSC, and this was a clever kludge to let some of the hres thru.

So thats how 1280x960 turns into 1280x480.

I can't say the explanation is perfect, but it sure as hell more logical than Sony, Panasonic, Bosch, Samsung, Pelco, Hikvision, Dahua, Honeywell, Vitek, all competitors mind you, agreed to just stretch every picture the same way and bald-faced lie about it, and without even a shred of technology to back it up! Collusion? Even when misleading with Sensup exaggerations, watch how carefully these companies (the big ones esp.) avoid telling direct lies. Not here. They are quite unequivocal about having increased resolution, clearly committing fraud if they are lying. Really...

IF YOU READ NOTHING ELSE READ THIS

I have asked kindly for the screeshots that were part of the "official" test. Since I have not received them yet I would ask you simply to do the following yourself as undeniable proof of higher non-stretched hres.

  1. open the 1280x480 screenshot in photoshop
  2. look at any horizontal line to verify that the pixels are unique and not merely stretched
  3. goto View, pixel aspect ratio correction and create a custom PAR of .5
  4. apply the PAR

Now imagine you are looking at the image for the first time without any knowledge of the technology used to create it. You verify the hres. The aspect ratio is correct. Nothing is stretched. What do you SEE wrong with the image?

thx to my better half for transciption and editing :)

"it would be great to have the vertical res too, but it just not possible with NTSC."

Jim, thank you for admitting as much, which takes us back to my statement you previously disagreed with, "The problem is it is transmitted by NTSC / PAL, which is the key limitation."

I get it Jim, you want us to double the size of the pixels to un-distort what is being transmitted across NTSC (the 1280 x 480). But that does not change the fact that 1280 x 480 is inherently distorted relative to reality and that showing or using 1280 x 480 (like Samsung actually does) is distorted. And doubling the size of the pixels does not magically add the details that are fundamentally missing because, as you admit, NTSC cannot deliver the vertical resolution.

Finally, as for images, Ethan has the weekend off. Now, it is Monday morning, he will read your dozen plus comments and respond appropriately, including with any relevant images.

Here's a comparison of the 1280H camera captured at 4CIF, the 1280H camera captured at 1280x480 but stretched vertically to 1280x960, and the 720p camera. Click it for a bigger version.

The 1280H camera captured at 1280H is a bit better than the 4CIF capture, but nowhere near the quality of 720p.

This brings up the two practical points:

  • 1280H is captured at 1280x480 in the Samsung DVR, which is heavily distorted. It's shown that way in screen shots in the web viewer and live view in SmartViewer.
  • If they do decide to simply capture at 1280x960, there is this potential quality advantage, with a proper aspect ratio and no distortion. However, then you're recording megapixel pixel counts without real megapixel resolution, which is not only impractical, it's completely deceptive to users.

Thanks kindly for the comparison matchup. The 720p is a good deal better than the 1280x960. And would be expected to be, since it has 1/3 greater number of distinct pixels. Likewise I would say the 4CIF to the 1280x960 is a good deal better too (zooming in makes it even clearer), though not as a dramatic improvement.

When you say its captured at 1280x480 in the DVR, are you saying on actual DVR/web client video playback (NOT file screenshots) it is distorted? Is this shown somewhere in that video?

However, then you're recording megapixel pixel counts without real megapixel resolution, which is not only impractical, it's completely deceptive to users.

When you say 'impractical' wouldn't that depend on the retention period? Are y'all also basing that on the rule that the file is double the pixels so must be double the bytes? I was thinking that if you do you scaling by just duping each line that the compressed file size won't increase too much. Cuz compression just loves when stretches of adjacent pixels have the same values. Im no expert but in photoshop with bilinear resampling the file went from 147K to 187K, maybe about a 25% increase, but not 100%. Hypothetically speaking, if someone decides to lower their compression level and that increased the storage by 25%, would that be 'impractical' too?

As for the deception part, since their not yet capturing the bloated file, wouldn't it be more fair to say 'deceptive if Samsung chooses not to clearly indicate the ramifications on storage', instead of just assuming they won't mention it?

P.S. I didn't see the 1280x960 image in the .zip, what was the jpg name?

thx again.

Apologly if I may look too simple and ask dumb question, because i almost got one of these ordered.

You say sometimes it stretches, sometimes it dont.

I try to keep simple. If i just buy this dvr and some bunch of cameras and nothing else.

And i just only will use monitor on dvr, when will i get that distortion ugly stretching and when dont i?

I only need watch live and playback video, like normal men.

If you view the local interface of the DVR, live and playback video will either be 4:3 or 16:9, depending on what type of monitor you use.

If you use the DVR web interface, the only time you will see the distorted image is if you use the screenshot button. Live and playback are shown in 4:3 viewing windows, regardless of what capture resolution is set on the DVR.

If you use Samsung SmartViewer (PC client software), you will see distorted 1280x480 video on live and playback.

Ok thx Ethan. I try just use DVR.

But it would still get some extra horizagonal resolution, ok?

I have 1600x1200 monitor, of the top off my head calculation give 4:3.

Hey i just thought something great for me, too!

I live in PAL-land. (near but not poland)

Thus then my citizenary right to get 576 vertically lines not just 480 as you are stucked with.

Th. you!

I only need watch live and playback video, like normal men.

You've come to the wrong place. None of us are quite normal....

Normality is a paved road;

It's comfortable to walk, but no flowers grow on it.

- Vincent Van Gogh

...camera is encoding reality at twice the width of reality.

Ok Honovich, I admit you baited me - this statement more than anything made me renew my subscription. Well done! :-()

Now to your metaphysical statement: I would say that would be impractical at best, 50 foot CMOS chips are not cheap either.

But with the exception of macro photography and microscopes, all lenses present a minature version of reality. Chips take that information in and sample it. But they are not informed about the true size of things in nature.

Don't worry though, I think I know what you are saying, that they are distorting the perspective. But though you may be most familiar with sampling in both the x/y axis at the same rate, nothing precludes asymmetrical sampling. And as long as the sample rates are adjusted before display, all will be in proportion.

But if you view this stream before it is fully reconstituted then you get a picture that has been compressed but not yet decompressed. What would you say to some viewing a RLL encoded file before decoding and then complaining about it being distorted? You would laugh, am I right?

Since it sounds like it comes with software that lets you get full use from this Dvr, minus the prt scr (which is trivial to resize), you should concentrate on the end result more. Maybe start with an HD camera that doesn't cost three times as much.

Great discussion, though.

Richard, thanks for chiming in!

"Since it sounds like it comes with software that lets you get full use from this Dvr"

The practical problem is that Samsung displays the distorted 'view' in many different parts of their clients. It's hard to get away from it, the way they have implemented it.

Since they are capturing at 4:3, they should just always display this as 4:3. I think we can agree on that?

It won't make it HD quality, but it will eliminate the distortion, which is what makes it really problematic.

Looks like they have their work cut out for them on those interfaces before they can call this Beyond anything. But if your nudge doesn't make them listen, maybe an inundated Technical Support will. >:(

John

I think I now understood what you were expecting from that camera.. Were you expecting to see a non-strectech image wich has an image ratio of cinema for example (21:9)? In that case you would have a larger horizontal field of view

I think most everyone would expect video to be displayed undistorted, regardless of how it is being processed or transmitted.

Howdy Ethan, if you are out there!

If its not too much trouble, would it be possible to get them actual screenshot files y'all made from the video? The 1280x480 and the 1280x960 ones. I'm interested in seeing if the 960v one has each horizontal line duped among other things.

Much obliged.

Hey Jim, here is a .zip of screenshots of the 1280H camera at both 1280x480 and 1280x960, taken using the screenshot function in Samsung SmartViewer. Note that SmartViewer, as discussed, displays video at 1280x480, with no option for 1280x960 live view, though.

Ethan, question for you: How come the Vertical FOV for the 1280x480 camera is so much bigger than the 720p?

1280x720

1280x960

So is the reason because this is a HD camera test by golly, not an MP camera test? And so the 5003 has a 4:3 sensor cropped down to 16:9? If so that's good news for the Samsung 1280H right? Since, correct me if I'm wrong, that means if it was not cropped then the FOV's would be more or less about the same right?

And that would mean, correct again please, that the Samsung 1280H would compare just as well against a full frame 1.3 MP Samsung as it does against the cropped frame one.

by any chance did you snap a uncropped pic with the 5003, i'm interested how it would handle the exposure if them bright lights were added into the frame.

Thx. Urgency=0

"How come the Vertical FOV for the 1280x480 camera is so much bigger than the 720p?"

The 1280 x 480 one captures a ~4:3 scene in reality, distorting / squishing / squeezing it down into a ~8:3 image. This has been the center of the debate for the last 20 comments and the point we are trying to explain to you.

The 720p one (1280 x 720) captures a ~16:9 scene at ~16:9 correctly and is therefore not distorted and needs no adjustment.

John or Ethan, is NTSC or PAL selectionable on device or do you order a new one? PAL should on the theory be much better true agree?

1280x720=hd

1280x575=?d

Maybe no good for you guys but will PAL be better than the NTSC?

I'm thinking that 960H and 1280H has its roots in Pal+. A lot of similarities. After John's first blog post about 960H, I ruled it out immediately. In fact, after reading this blog post, I came close to clicking out of the post and getting on with my life. However, I decided to download the images and examine them myself in Adobe Photoshop CC. After doing this, I saw a few things that took me by surprise and some things that could be improved upon if it was not for current limitations. I would not rule this Analog Technology out yet. My instinct and intuition, along with the credibility of companies investing in this tech, tell me that we could see something that could be noteworthy within the next couple of years.

Howdy! More for my own understanding than anything, I did a little mockup of the functional process of the H technology, based on the information presented so far, as well what I got out of Vlado's latest update 'CCTV - From Light to Pixels', where he gives his take on 960H. I tried to keep the pictures small as default, but there's plenty more detail if you zoom. Nothing shocking here, but hopefully this will help. If I've made any mistakes, y'all don't be shy!

The basic idea was to create a 1/10 resolution functional example of the transformations thru the image pipeline. At this scale you can get a real idea of what happens at the pixel level. To make the math easier I'm rounding the actual sensor pixel count from 1305x1049 to 1280x960, The process is the same, though either way.

1. So 1/10 of 1280x960 is 128/96, and the empty grid below represents the mock sensor before exposure and each sqaure a mock 'pixel'.. Our grid has a horizontal resolution of 128 and a vertical resolution of 96.

2. The whole sensor is exposed to the FOV and the pixels read out. Now any two adjacent pixels are more than likely to have different values and so the true resolution is 128x96.

3. On account of the fact that real NTSC has a hard limit of 480, we can only use 1/10 of that or 48 lines in our mock NTSC. Since we are now at 96 lines, we have to reduce by half the number of vertical lines in order to comply with the standard. To do this we average the pixel values between the first two horizontal lines and replace both of them with two new identical lines both containing the averaged values. Note that even though our resolution is technically still 128x96, our real resolution is 128x48 since every two vertically adjacent pixels have the same value. Also note the substantial loss in detail in the image.

4. Next we reduce our image footprint by half simply by removing the duplicate horizontal lines. Now we have a 128x48 image ready to fit in the NTSC pipe to send to the DVR.

5. Finally, on the other side of the pipe, the DVR readies the image to be displayed. It uncompresses the image by duplicating, one horizontal row at a time, the stream, until we have a 128x96 size image. Although in the end it actually has a true resolution of only 128x48. This recreates the exact image we created in step three, after we averaged the pixels vertically.

*this here is only a functional equivalence, for example one would think that steps 3 and 4 would probably be combined in one action, and there are likely other optimizations.

"the DVR readies the image to be displayed. It uncompresses the image by duplicating, one horizontal row at a time, the stream, until we have a 128x96 size image."

Right, except that Samsung, in many parts of their clients, does not duplicate rows, choosing to display it in a distorted 128x48 mode.

As we've discussed repeatedly, displaying it always in a 4:3 aspect ratio will eliminate the distortion problem. It will not make it HD or HD quality because of the limitations you describe above, but it will at least no longer be distorted.

JIm, you're an obsessive compulsive after my own heart. If it weren't for the obsessive compulsives in the world, we wouldn't learn anything.

Here's is a question I feel must have been addressed somewhere, though I don't see it scanning down.

What happens if this camera is hooked up directly up to an NTSC monitor? Is it in proportion? Is anything cut-off? Are there more measurable TV lines? What if the monitor is one of those B&W 950 tvl pro ones? Not that I think people are likely to use it that way, but just to get an idea of the voodoo going on.

Hi Richard, when it's hooked up to any standard NTSC device and captured at a non-distorted aspect ratio, it delivers slightly more detail than 960H or 900TVL cameras, but it's marginal. Nothing is cut off.

See this comparison, with the camera captured at a normal aspect ratio:

As far as how it would look on the 950TVL monitor, I don't know. I'm guessing again, marginally better than some analog cameras, but I can't say for sure.

Ethan, the images seem a little strange to me. The 900TVL camera appears to reproduce the eye chart a little better than the 1280H or 960H cameras but the 1280H camera appears to reproduce the license plate slightly better than the other two. Why would that be?

Another better real world question: (and obvious for all who try mix/match camera to their misfortunate legacy equip.)

What would be seen (screenview) if 1280H camera goes on into a competitor DVR, instead?

Right away my gut said 'flattned down picture'.

But then voice said 'no it will be square', (because i'm wrong more than right, on the average). This is good and fair question for all to know the answer.

E,

I can't vouch for either SD DVRs or 1280H cameras but we are currently installing 960H cameras to replace older SD analog cameras and we get 4:3 SD images. Our current encoders are IndigoVision 4SIF and any analog cameras we've connected to them deliver basically the same images, no matter what their stated definition is. We've tested three different 960H cameras: inMotion, Vitek and Everfocus and all deliver the same resolution (but not necessarily the same image quality in differing conditions).

We also observed the same results with other encoders during our system evaluations in 2012/2013.

What I've found is that true SD analog cameras are a dying technology. Other than super cheap no-name cameras, it appears that much of the industry is switching to 960H and higher technology. I believe that the main impetus for the switch is profits. We were paying around $100 each for analog SD box cameras but have found that most 960H box cameras sell for closer to $200 - double the old price. Domes are also priced proportionally higher.

And we haven't observed much difference in picture quality or features to justify the huge price difference. I assume that to obtain any benefit from these so-called higher resolution cameras, it is necessary to use appropriate encoding/decoding. The same problem applies to analog and digital HD technologies like AHD, HDCVI, HD-TVI and HD-SDI.

Y'all, is anyone still of the opinion that Samsung is simply taking an 704 SD image and resizing it to 1280 by stretching it out, meaning making it bigger without adding any detail? If no one believes that then shouldn't statements like them below in the report be fixed, so new people reading it don't get misled?

It is obvious that Samsung 1280H stretches the image / video extremely wide...

Those of you who have had the misfortune of using 960H will immediately notice it is the same absurd trick, simply taken to the next level by Samsung, stretching even wider.

The 1280H resolution stretches the image by 81.8% horizontally...

From SD to 1280H, it's a 'stretch' from 4:3 to 24:9 or closer to 100%

But the question remains - what is the worth of taking lower resolution images and scaling them up?

In this case, the video / image is scaled up from 704 x 480 to 1240 x 480.

Related Reports on Low Light

Axis 4MP Camera Tested (M3046-V) on Sep 28, 2016
Axis has brought 4MP to its camera line in the new M3046-V, the highest resolution model in their revamped M30 series. We bought and tested this...
History of Video Surveillance on Sep 22, 2016
This is a concise history of video surveillance covering the past decade.  The goal is to help professionals newer to the industry understand...
Camera Course September 2016 on Sep 15, 2016
This is the only independent surveillance camera course, based on in-depth product and technology testing. Lots of manufacturer training exists...
Hikvision 4K Camera Tested on Sep 09, 2016
Hikvision is the most common choice for low price entry level products but they are also competing with low light models, smart CODECs, WDR...
Pelco Optera 270° Camera Tested on Sep 06, 2016
Multi-imager cameras are typically 180° or 360°. Pelco has released a fixed 270° versions of their Optera intended to cover exterior building...
Panoramic Camera (Fisheye / Multi-Imager) Guide on Aug 31, 2016
Panoramic cameras, including fisheyes and multi-imagers, have become increasingly widespread, with most manufacturers offering fisheyes and an...
Dahua Smart H.264+ Tested on Aug 17, 2016
The smart codec trend continues, with Dahua adding Smart H.264+ to many of their models in new firmware. In this test we see how it performs and...
21 Scenes Lux Readings For Video Surveillance on Aug 11, 2016
IPVM went to numerous locations and measured the lux readings / light levels at each one. Here is an example of the approach we took and output...
Vivotek 12MP IR Multi-Imager Tested on Aug 09, 2016
A key weakness of multi-imager cameras is poor low light. While integrated IR has become dominant in cameras overall, it is still uncommon in...
Integrated IR Camera Use Dominates - 2016 Statistics on Aug 03, 2016
In 2014, IPVM statistics showed integrated IR camera use was surging. Now, new IPVM statistics show that the surge continues, with integrated IR...

Most Recent Industry Reports

Camera Calculator Class and IPVM Member Orientation October 2016 on Sep 28, 2016
Members, learn how to better design video surveillance systems and get the most out of your IPVM memberships with 2 upcoming live classes. Both...
Axis 4MP Camera Tested (M3046-V) on Sep 28, 2016
Axis has brought 4MP to its camera line in the new M3046-V, the highest resolution model in their revamped M30 series. We bought and tested this...
Hiring Spree At Aimetis 6 Months After Being Acquired on Sep 28, 2016
Aimetis was acquired in April 2016, and is now expanding almost all of their departments, hiring employees from Axis and other industry...
Camio Natural Language Processing Tested on Sep 27, 2016
The ex-Googler led team from Camio has advanced its video monitoring offering to include natural language processing. Camio ingests video,...
Access Door Controller Configuration Guide on Sep 27, 2016
Properly configuring access control door controllers is key to a professional system. These devices have fundamental settings that must be...
Hacked Dahua Cameras Drive Massive Cyber Attack on Sep 27, 2016
Cyber attacks are accelerating and IP cameras are behind many of them. Worse, last week, a 'massive' attack was carried out using numerous Dahua...
Axis Secretly Paid Anixter Sales People To Push Axis NVRs on Sep 26, 2016
Internal Axis communication shows how Axis paid Anixter and Tri-Ed sales people with secret bonuses to push Axis NVRs. In this report, we examine...
VLANs for Video Surveillance Tutorial on Sep 26, 2016
Many people confidently say to 'use VLANs' as an answer to IP video networking problems and as a way to signal expertise. But how should VLANs be...
Ambarella CEO Admits H.265 and 4K Not Popular on Sep 26, 2016
Ambarella is the main chip provider for high-end surveillance cameras driving higher resolution and new CODECs. While Ambarella has been pushing...

The world's leading video surveillance information source, IPVM provides the best reporting, testing and training for 10,000+ members globally. Dedicated to independent and objective information, we uniquely refuse any and all advertisements, sponsorship and consulting from manufacturers.

About | FAQ | Contact