The appropriate standard to describe IP camera resolution

[ting.han 0]

Hi,

 

I was looking at some IP camera specifications and I came across mainly the two ways of describing their resolution. One is using the HTVL, which in my opinion should be used only for analog cameras. Another standard is pixel. Sometimes a camera would have both on their spec sheet.

 

What is the right way to describe the resolution of a digital camera? Why are HTVL used in some cases? Please share your thoughts.

 

Thanks for your time.

[Soundy 1]

If by "HTVL" you mean terms like 720p and 1080p... there is no valid reason to use them other than for marketing. Technically, even "MP" designations are often only approximate - for example, 1280x1024 is typically sold as "1.3MP" even though it's technically 1.25MP.

 

The only truly accurate measure IS the listed HxV resolution.

[jhonovich 0]

Good topic. IP camera manufacturers essentially stopped measuring resolution using HTVL. Its replacement is the pixel (as Soundy mentioned).

 

Using pixels as a measurement is problematic as pixel measurements can distort the actual visual resolution a camera delivers. This is a common criticism in the digital photography world that more pixels does not always equal better resolution.

 

The reality is HTVL has been abandoned and it's a loss for the industry.

 

There's some discussion on IP Video Market about these topics, including:

 

http://ipvideomarket.info/report/megapixel_cameras_better_image_quality

 

and

 

http://ipvideomarket.info/report/how_well_do_ip_cameras_work_in_low_light

[Soundy 1]

Good topic. IP camera manufacturers essentially stopped measuring resolution using HTVL. Its replacement is the pixel (as Soundy mentioned).

 

I wouldn't call it a "replacement" really - the pixel IS the proper unit of measurement for an electronic (CCD or CMOS) image sensor, after all. "TV Lines" is a throwback to when tubes were used for imaging and didn't have individual pixels.

 

Using pixels as a measurement is problematic as pixel measurements can distort the actual visual resolution a camera delivers.

 

Unless they're non-square pixels, how do you figure?

 

This is a common criticism in the digital photography world that more pixels does not always equal better resolution.

 

This is true with any type of cameras, though - there are too many other factors that come into play that are often overlooked, even with standard analog CCTV. Pixel size plays a huge part in the sensor's ability to gather light, and thus in its low-light performance. Lens optical quality is regularly discussed and debated in photography circles, but generally taken for granted with CCTV (one of Avigilon's advantages, in addition to a larger sensor, is the ability to use high-quality photographic lenses).

 

The reality is HTVL has been abandoned and it's a loss for the industry.

 

Why is it a loss? It's a limited, inaccurate term. It will tell you the vertical resolution of a sensor, but not the horizontal resolution. Trying to carry over analog terms into an all-digital world is just folly; the sellers may like the ability to snow buyers with meaningless "specs" but it certainly doesn't benefit the end user.

 

Might as well advertise a car by touting its ability to fire up in only three turns of the crank handle.

[jhonovich 0]

TV lines are an expression of actual resolving power of an image and as such represents something meaningful to a viewer.

 

Pixels are a physical element of a sensor and are, at best, indirectly related to actual visual resolution.

 

I think pixels are misleading as discussed at length in the articles I cited.

[Baylab 0]

TV lines are an expression of actual resolving power of an image and as such represents something meaningful to a viewer.

 

Pixels are a physical element of a sensor and are, at best, indirectly related to actual visual resolution.

 

I think pixels are misleading as discussed at length in the articles I cited.

 

Actually, even two sensors have same pixel number and same pixel pitch, and used with same lens with same focusing, you still can get different result.

 

for the high resolution sensor, especially the one with small pixel, the crosstalk between the neighboring pixel is quite significant (light will leak from one pixel to another pixel). such kind of crosstalk will degrade the image quality definitely.

[shropna 0]

TV lines are an expression of actual resolving power of an image and as such represents something meaningful to a viewer.

 

Pixels are a physical element of a sensor and are, at best, indirectly related to actual visual resolution.

 

I think pixels are misleading as discussed at length in the articles I cited.

 

Actually, even two sensors have same pixel number and same pixel pitch, and used with same lens with same focusing, you still can get different result.

 

for the high resolution sensor, especially the one with small pixel, the crosstalk between the neighboring pixel is quite significant (light will leak from one pixel to another pixel). such kind of crosstalk will degrade the image quality definitely.

I don't think so... Crosstalk between pixels "possible" in over lighted CCD, but not in CMOS. Different real resolution, but not image size in pixels caused more optical and little bit compression effects...

[Baylab 0]

TV lines are an expression of actual resolving power of an image and as such represents something meaningful to a viewer.

 

Pixels are a physical element of a sensor and are, at best, indirectly related to actual visual resolution.

 

I think pixels are misleading as discussed at length in the articles I cited.

 

Actually, even two sensors have same pixel number and same pixel pitch, and used with same lens with same focusing, you still can get different result.

 

for the high resolution sensor, especially the one with small pixel, the crosstalk between the neighboring pixel is quite significant (light will leak from one pixel to another pixel). such kind of crosstalk will degrade the image quality definitely.

I don't think so... Crosstalk between pixels "possible" in over lighted CCD, but not in CMOS. Different real resolution, but not image size in pixels caused more optical and little bit compression effects...

 

That phenomena happen in CCD is called "blooming", it is another story ( overflow of electric charge VS light leakage) .

 

there is no totally opaque wall between neighboring pixels (so researchers proposed to add aluminum (or other reflective )wall between pixel) . the colour sensor is worse due to the bayer filter.

[hardwired 0]

Until one of us steps up to the plate and purchases a true analysis package for this purpose (http://www.imatest.com/home, or something like it), all of our observations about resolution will only be subjective. Granted, some differences are more easily visible than others, but the effects of different processing and compression schemes often create more effect on the final image usability as does differences in the native image quality coming off the imager.

 

Machine vision camera manufacturers regularly provide a wealth of information about native image sensor sensitivities, resolutions, etc. that CCTV manufacturers do not, because their users demand it.

 

Among other CCTV market problems, manufacturers that do not inflate their standards risk being compared poorly to those that greatly exaggerate their specs, because less discerning buyers will make purchasing decisions solely from those specifications, rather than any objective comparisons.

 

I think that the greatest advantage of the "lines of resolution method" is that it did allow the observation of a whole package (lens, camera, and compression) in operation. Pixel-based approaches are lacking in that regard.

 

I can have a nice looking 2048X1536 pixel image that I can make out great detail, faces, plates, etc.. or I can have a 2048X1536 pixel image with the iris shut too far, compression at maximum, and out of focus.... Still the same "resolution" image. We need manufacturers to comply to an industry standard (anyone expect that anytime soon?), or a reasonably effective test bench configuration and testing methods, to really objectively compare cameras.

 

Who among us is going to step up to the bench, and help hold manufacturers feet to the fire on this subject?

[Baylab 0]

imatest is a good choice the evaluate the quality of image. there are some free software can do similar job.

 

ImageJ, developed by people in NIH, should be the most suitable software to do this.

 

http://rsbweb.nih.gov/ij/download.html

 

of course, you need another Plug-in called SE MTF, which is free too.

 

http://rsbweb.nih.gov/ij/plugins/se-mtf/index.html

[shropna 0]

imatest is a good choice the evaluate the quality of image. there are some free software can do similar job.

 

ImageJ, developed by people in NIH, should be the most suitable software to do this.

 

http://rsbweb.nih.gov/ij/download.html

 

of course, you need another Plug-in called SE MTF, which is free too.

 

http://rsbweb.nih.gov/ij/plugins/se-mtf/index.html

I'm still use CCTV Labs Test Chart

For example, Sanyo HD4000 test:

[shropna 0]

I can't post full resolution image there, but, of course, TVL is more informative than image pixel size....

[Baylab 0]

" title="Applause" /> this test result is quite impressive.