Tips on sharpness...

ccaillouet

Member
I'm a video engineer who has been working in high definition since 1987. I have seen at least six generations of cameras with a variety of lenses come along before the introduction of the 1/3 inch models. Lately, I have been shooting some tests under various conditions in different formats and have noticed the following regarding sharp pictures with the HVX200. Much of it applies to any small camera. Jarred thought some of you might like to read my conclusions.

I hope this is helpful.

What have you noticed in your world?

cheers,
crc
--------------

1) Regarding the small chip problem and aperture:
I find that i prefer the quality of the system around f4 - f5.6 because the farther open you get, the more you see chromatic aberrations, especially around the edges of the image. These show up as a red or blue fringe on edges. The diffraction limiting spoken about with 1/3 inch chips is not severe at these settings and the failure is a monochromatic softening of the image, which is the subject of this discussion in general and can be dealt with, to a certain extent. It's a balancing act to maximize sharpness and minimize artifacts, especially while trying to control depth of field. You have to decide which is more important for your application. Big screens blow up small artifacts but they also expose softness.

2) Focus:
Focus is the single biggest problem, especially at longer focal lengths. This lens is really pretty long for a 1/3 inch camera and focus can be an issue if the subject moves very much. An example is shooting animals from a distance. The low res viewfinder and LCD are limited, even with mono and EVF detail in the small one. You just have to keep checking to make sure focus is good. If you just blur one pixel into the next one, you have halved your highest resolution and it goes downhill from there, reducing detail contrast and apparent resolution.

3) Detail settings:
I usually like less detail in a camera but i find that the HVX needs detail set around "0". It is possible to shoot with a little less to minimize noise enhancement, and then to use the Unsharp Mask filter in FCP to bring it back. I use 50% at 2 pixels and a 5-10% threshold. For some reason, i found that the Luminosity check box (in the Unsharp Mask filter) needs to be off or the noise gets a little blocky. You can use variations of these settings to make some soft video "pop" a little. The bigger that you make the pixel value, the more of the "white edge" effect you may get. If you are shooting mattes, you want to minimize the Detail setting in the camera and do any sharpening after the matting process.

4) Gain:
Keep Gain at 0 to minimize noise enhancement. Detail enhancement aggravates noise. Noise uses bits in the compressor. Too few bits causes blocking artifacts to appear.

5) Record Format:
Don't bother with 1080i unless your client specifies it. The sensor arrangement and native resolution of this camera is well suited to 720P, which is also where the variable frame rates work, and the down-conversion to 720P has the effect of doing a little noise filtering in an area where there is little, if any, detail. You can shoot 1080 if you want to, but, in my opinion, it won't buy you anything except marketing advantage. And if you are shooting at 24PN, it makes a big difference in record time, AND transfer time per recorded minute. 1080i60/24P uses the same record resources as any other 1080i.

6) OIS:
Don't use the OIS unless you really need it. Use a heavier tripod. The OIS is good but it can cause small movements which can add just a little blur at normal shutter speeds. Remember, a one pixel blur can reduce the resolution by HALF and you don't need that. Also, using a wireless or wired remote will prevent bumps when recording still shots and helps when using the Pre-record feature - good for nature recording.

7) Exposure:
Work hard to use the bits that you have by using good exposure control. Unless you are looking for a special dark look, full exposure will usually work better for this camera. Sharpness is about contrast and is more important at lower resolution because the vision system is more sensitive to contrast at about 1/10 of the limiting resolution. So for most viewing distances and most scenes, good contrast at 200 lines will make a scene look sharper than limiting resolution at 1000 lines. Of course, more expensive cameras and lenses will give you more of both ends, but we are dealing with what we have available. Focus and exposure, along with lighting and subject content, determine contrast.

8) A few more words on noise:
Full exposure will also reduce the visibility of noise in the picture. Camera noise comes in different flavors. Some of the noise in the sensor does not vary with with the signal so a larger signal masks the noise while a darker signal looks noisier. The automatic gain functions in the eye/brain also kick in to pull the noise up in dark scenes. This is another reason to keep the signal higher if you can, while still getting the look that you want. Try using contrast within the scene to give the impression of darkness while keeping the eye from opening up, and to attract the eye where you want it. Also, playing with different gamma and pedestal settings can move the noise into an area that does not get amplified so much. The Normal HD gamma setting has a 4x amplification at the bottom few percent of the curve. That's how gamma works - darker signal, more gain - and the 4x limit is to keep the picture from getting even noisier. The CineLike settings have a toe that suppresses the black gain a little and might help you keep the noise down.
 
Fabulous info!

I am burning your tips into my brain, and will know what to do when my HVX comes.

Thanks.
 
One more comment on exposure:

This comes up when shooting highly saturated colors.

In the high definition matrix equations, blue contributes only 11% of the luma value, red gives 30% and green, 59%. This means that if you had a 100% saturated blue, you could only get a luma value of around 15% before the blue channel hit clipping. Now, this is a pretty severe example but you should be starting to get the message. If you use the zebra, the marker or even luma on a waveform monitor, you could end up wondering why there is no detail in saturated colors, even greens. That's why we use an RGB parade waveform whenever possible. Color difference parade can also show you clipping, but it's not as obvious.

So you have to adjust mentally to backing off on the exposure if you want to hold detail in saturated colors - yep, it's another balancing act. Good video is full of them...

cheers,
crc
 
It's too bad Panasonic had to go for 1080. In my testing, there is a definate difference between footage shot at 1080 and footage at 720 (at least 50 lines h & v.) I would have liked the lower bandwidth format without taking the resolution hit.

Great tips, by the way. :thumbsup:
 
Thanks, great info. This deserves a sticky.

One question I posted in another thread about range. If I use a spot meter and set it at ASA 320 (so all measurements are the same). Make the key of my scene f5.6 then what range do I have on either side? If my deepest shadows are f2.2 am I going to end up with the red an blue fringes there? and how far can I push the highlights before I my picture degrades, F11 maybe?

Thanks again for taking the time to share your knowledge and findings.
 
Gopher_Greene said:
If I use a spot meter and set it at ASA 320 (so all measurements are the same). Make the key of my scene f5.6 then what range do I have on either side? If my deepest shadows are f2.2 am I going to end up with the red an blue fringes there? and how far can I push the highlights before I my picture degrades, F11 maybe?

I'm not sure if i understand what you are asking. The issue of chromatic aberations (the red and blue fringes) is related to the lens aperture setting, not the amount of light in the scene. You might see some color fringing on out-of-focus objects with cheap lenses, but the problem that i am talking about is on in-focus objects when a lens is near wide open, especially near the edges of the frame. If you set the lens to "Open" or f2, you might start to see some fringing. How big you magnify the picture and how picky you are will determine how much of a problem it is for you.

The same is true for diffraction limiting. It happens when you stop any lens down past a certain point. The issue for us is that physics limits lens performance on small sensors to the point that we start to see some softening past about f5.6. Again, how much it bothers you depends on your application. If you like a soft look, you might think it is not a problem. If you like snappy, video-look pictures, you may turn the Detail setting up and not see a problem at f11. On the other hand, you might be picky, blow the images up a lot and not like what you see at f5.6. In any case, the effect is an overall softening, not color fringing, so, to my eye, it is not as distracting and is easier to correct in post, up to a point.

Now, you might cause other artifacts, like sharp clipping edges, if you push the highlights too far, especially in one of the video gamma modes, but these are not usually colored. They can, however, be one giveaway that you are shooting video. Film doesn't fail that way. Use one of the CineGamma modes to avoid the problem.

Different strokes...

cheers,
crc
 
Pascal_Parvex said:
So, the 1080p mode is just PR?

No, it's really there and it is progressive.
It just doesn't give you the space advantage of recording in 720P24PN.
How much additional resolution it gives you depends on the care that you take getting it, up to the limits of the camera and the format. My personal preference is to use 720P but my work is on small to medium venue displays. 720P works fine for me so it makes sense all around.

If you have a need for 1080i or P, the camera will give you the format, but the tradeoffs are there, just like in the bigger cameras. You get more resolution and bit depth if you record outside of a Varicam or a CineAlta. You get more resolution from a bigger sensor and better lenses. The questions are 1) how much do you need; 2) how much does your client think you need; and 3) how much is the production willing to pay for, in money and in time?

Test it yourself. Everyone has an opinion.

cheers,
crc
 
ccaillouet said:
One more comment on exposure:
In the high definition matrix equations, blue contributes only 11% of the luma value, red gives 30% and green, 59%.

This is slightly off the subject of the HVX but it affects everyone starting to work in HD.

If you were paying close attention, you might have noticed that i quoted above the numbers for SDTV, not HDTV color. (Hey, i can blame it on getting old. What's your excuse?) The HD situation is even worse: green, 72%; red, 21%; blue 7%. This is from ITU-R Rec 709.

Who cares, right?

This might be important to you if you are working with software that advertises that it can handle HD images, without properly dealing with the color. I ran into this while passing information among earlier versions of Final Cut Pro, AfterEffects and VideoScope on the Mac and Pixeltools on the PC. Even though each of them supposedly handled Quicktime movies, they obviously made different assumptions about the color encoding equations. I ended up having to apply special filters to get my clips to encode anywhere close to the original color and luma values.

More recently, i found similar problems between FCP and Sorenson Suite 4. Evidently, many software vendors think that they just have to scale up the algorithms and not deal with different color equations.

None of this has much to do with "color space". The color primary definitions are somewhat different, but the effects of using the wrong equations are much broader than not reproducing a few colors. This affects the GAMMA of the luma signal. If you don't believe this, look at HD color bars in black and white and notice the big jump between the luma of the middle two bars.

This is a big problem and no one even mentions it.

So if you are having problems moving clips among applications or across platforms, you might be the victim of sloppy transcoding. Try a different path and see if it works better.

BTW, this applies to set-top boxes as well. If you happen to be encoding MPEG2 for kiosk or other server playback, you might find that some encodes neglect to set all the flags correctly and you could end up with the MPEG2 decoder screwing up the color as well. It's probably just doing what it thinks it's being told.

Ain't the multiformat world exciting?

cheers,
crc
 
Back
Top