HV20: Cinemode Softness (loss of image detail) Pic

[Eugenia Loli-Queru]

>8) A good compromise setting for 24P "film" shooters who want the extra detail might be Tv 1/48th, NEUTRAL image processing,

What do you mean by "neutral"? How do you get that?

[Thomas Smet]

You guys are putting way too much on this. It isn't even an issue really. Why do you think nobody noticed it before? Because you are all making this out to be bigger then it really is. I mean look at how you are testing the images by putting them side by side. If you have to look that close to see the difference then is anybody just watching the footage going to notice anything?

You are all making this out to be some kind of problem when it isn't. The raw tap from the chip without the electronic sharpness is going to be soft no matter what camera you use. The reason why 4K is desired is because you can have the raw feed without any sharpness added at all and still have a clean image.

Canon actually had the vision to think what film makers usually want and that is a clean but soft image. In my opinion super sharp images are a curse of the computer game and flash generation who think video should be as crystal clear as a flash animation. If the HV20 didn't have a normal mode nobody would have ever had a problem with the softness of the Cinema mode.

Give me an image from a live HDMI capture and I will try to give you a result close to what you can get in a normal mode. I cannot beleive out of all the details in the image the tiny window in the background is going to be pointed out that the cinema image cannot be sharpened as much. Did it ever occur to anybody that the reason the window may be softer is because the DOF is slightly different for whatever reason due to a change in the shooting modes? Clearly that window is more of a focus blur and not jut raw detail.

There is no reduction in detail going on. There is no magic chip in the camera to perform a blur on the image. If there was it would be the first camera or hardware chip I have heard of to blur the image. Of course hardware blurring in the camera is going to be different then software blurring. Chances are the camera sharpens the video while it is still as a bayer pattern. The RGB image will be sharpened as it is created from the bayer pattern. This is something you could never do with a compressed HDV image that is already RGB or YUV. This is where you will still notice a little bit more detail in a sharp image from the camera and doing the sharpen yourself. If this slight edge in detail is your cup of tea then hey thats your thing I guess. If you want more detail with better control go out and buy a Cinealta F900 and quit trying to whine about a $1,000.00 camera not being perfect.

[Stu Maschwitz]

Amen Thomas! With much gratitude to those who have shared their test results with all of us, it's back to work for me...

-Stu

[Joseph H. Moore]

What do you mean by "neutral"? How do you get that?

In the "IMAGE EFFECTS" menu. Third option, after "VIVID."

[Chris Barcellos]

I'm shooting another 48 Hour Film Project this weekend, using the HV20 with a Letus Adapter. In the last one, I shot in Cine mode. My feeling at the time was that color correction would be easier... Turns out we didn't have time to color correct, so this time I am going with TV setting, shutter at 1/60-- since the 48 Hour Film Project says " no 24p. I will let you know my impressions after that. I expect to lose some shadow detail shooting this way....

[Stu Maschwitz]

Originally Posted by Chris Barcellos

...since the 48 Hour Film Project says "no 24p."...

What? Then they should call it the 48 Hour VIDEO Project!

-Stu

[Eki Halkka]

Originally Posted by Stu Maschwitz

Amen Thomas!

What he said ;-)


...Even though i agree that this is kind of silly nitpicking, I shot my own quick tests, these should clarify a bit which kind of artifacts we are dealing with when working with different settings. It's all a matter of picking what's important to YOU personally, and doing the best you can with that.

My workflow was the following: i took the raw .m2t to After Effects, and exported .psd files of all the frames i needed. On some of them, i did some simple curves adjustments and unsharp mask sharpening.

There's no additional compression in these images, only the HDV compression (I'm pretty sure Eugenia's "lossles video compression" is not really lossless) and of course the lossless .png compression of the final cropped images (crop to more manageable size was done preserving pixel per pixel detail).

Some detail has probably been lost when converting to RGB, but that should be minimal, and affect all the images the same way.

Here we go.

First, here's TV mode image at automatic settings. Highlights are blown out, and there's some clearly visible sharpening artifacts:

http://eki.pp.fi/temp/Eki/HV20/Eki_tv_auto.png

Here's TV mode exposed for highlights. the sharpening artifacts are still there, but there's good details in the highlights. Unfortunately, shadow areas are quite dark...

http://eki.pp.fi/temp/Eki/HV20/Eki_tv_highlights.png

Here's Cine Mode at automatic settings. Some of the highlight detail is blown out, but we can see quite well in the shadows. Contrary to what i thought, there seems to be a tiny bit of sharpening also in this mode, seen i.e. at the balcony chair's legs near Otto The Cat's head. Also, when looking i.e. at the door details, it looks like the detail we get is about one pixel size - i wouldn't call that bad. The overall image looks less sharp than with TV mode.

http://eki.pp.fi/temp/Eki/HV20/Eki_cinemode_auto.png

Here's Cine Mode exposed for highlights. There's still some detail in the darker areas of the shot, and highlight areas are nice. This is how i would have taken this shot.

http://eki.pp.fi/temp/Eki/HV20/Eki_c...highlights.png

Next up: adjusted images

Here's the Cine Mode image with adjusted contrast only.

http://eki.pp.fi/temp/Eki/HV20/Eki_c...t_tv_gamma.png

Here's the same image, with unsharp mask sharpening. Overall, this is rather close to the TV mode, IMO. There's some lost detail, visible especially in the shadow parts of the curtain. Note that i wasn't trying to make the image look "good", i tried to make it look like it was shot in TV mode auto exposure (the same goal as in my previous example, Not Pretty But Similar).

http://eki.pp.fi/temp/Eki/HV20/Eki_c...ma_sharpen.png

Here's highlights-exposed TV mode image adjusted to look like Cine Mode image. The overall look is quite close again, but there's a lot of noise, loss of detail and compression artifacts in the shadow areas, and the sharpening artifacts rear their ugly head. To me, this is a major turn-off, i find the sharpening in i.e. the balcony chair horrendous.

http://eki.pp.fi/temp/Eki/HV20/Eki_t...mode_gamma.png

That's it - what's left is to pick your poison. *Lost detail in low contrast areas* versus *sharpening artifacts and noisy shadows / blown out highlights*.

It's all a matter of preference.

Edit: here's one more image - i took the highlight exposed Cine Mode image and the highlight exposed / curve corrected TV mode image, separated the luminosity part (lab color) and enlarged to 300% with nearest neighbour scaling. This comparision is of course a bit unfair for TV mode, because of the gamma adjustment which brought out the compression artifacts and noise. But it shows the compression-related loss of detail in cinemode, and the edge sharpening artifacts in TV mode quite well.

http://eki.pp.fi/temp/Eki/HV20/Eki_c...omparision.png

One more thing worth mentioning: all post sharpening in the examples so far has been done with simple algorithms, better looking results could probably be obtained i.e. with Virtual Dub's free warp sharp filter, or something like Focus Magic photoshop plugin (which is good but not free).

[Thomas Smet]

Eki thanks for that post.

As you can see when you sharpen the HDV cinema image it can get closer but it will never be the same. That is due to the fact that the camera sharpens a raw RGB image and that it sharpens it as it processes the image from the bayer pattern.

The realstream guys had the same issue with the modification to the DVX100. In that case however they were pulling a bayer pattern because of the pixel shift so they could work the sharpening back into the image during the image processing from conversion from bayer to RGB image. It really helped pull back in some sharpness.

To me your cinema image looks much better and more natural. It looks almost photographic although slightly soft. It really does down to personal choice though. Some people like me love the natural soft filmlike look of the cinema mode while others want the extra sharpness. Look at the comparison image for example. In photography there should be a natural transition between edges into the background. The normal image has too harse of an edge and isn't natural at all plus it creates a ring that shouldn't be there. For compositing the normal image will never have as clean of a key as a cinema image would. You will either have to live with a ring around your subject or mask the edges which already screams out fake composite. I learned from a compositor from ILM who told me that the edge should always look exactly the way it would in nature which means no sharpening or matte chokers. If you do have to use a matte choker make sure the edges look how they would if you photographed that subject in front of that background for real.

It needs to be said though that the cinmea mode is not killing or stealing detail. It is just that the normal modes add in a lot of sharpened detail which make the image look like it is more detailed. For a lot of consumers this is great because they may not want to sit around and adjust all their footage.

Again like I said however we really do have to give Canon props for having the vision to actually give film makers a choice on how they want their images to look. While the camera is aimed at the consumer market it is clear that Canon also wanted to tap into the budding film market as well.

So for any new people...

1. Cinema mode is softer which is the natural way it is supposed to be. It is not a reduction in detail but the lack of added electronic sharpness.
2. Normal modes are sharper because the DSP can work with the raw image to add sharpness to the raw image.

You choose which mode works best for you and which one gives the most pleasant look for you.

[Mathieu Kassovitz]

Originally Posted by Joseph H. Moore

Nope, it's about all of the above.

Lol

You've taken the words right out of my mouth. Above and below, all included with no exceptions.

Canon's cinemode is the best way to get it.

[Joseph H. Moore]

Thomas,
There is no "NORMAL" mode. ;-) That implies the CINE is an abnormal choice. There are several program modes, of which CINE is one.

[Thomas Smet]

Sorry I was just too lazy to write out all the other modes. :)

[Joseph H. Moore]

the 48 Hour Film Project says " no 24p.

You should be able to still shoot 24P and deliver it "telecined" in the 60i HDV stream. (That is to say, shoot in 24P mode, but edit it as 60i. Works great, keeps the feel, everybody's happy.) I think they just don't want to deal with the vagaries of a bunch of different formats.

[Thomas Smet]

There may be another reason for using Cinema mode for 24p shooting.

The reason why some people find 24p from digital cameras to not be as smooth as 24p from film cameras is due to the edge sharpness and DOF. unnatural sharp digital edges make 24p motion seem a little bit more jittery then film. It is a little bit more subtle effect something like how 24p animation with motion blur will always look better then 24p animation without motion blur. The soft natural edges of the cinema mode help transition each frame by not creating an abrupt sharp change in position. It is subtle but it can help.

Just another thing to keep in mind about the cinema mode if you plan on shooting 24p.

[Joseph H. Moore]

It needs to be said though that the cinmea mode is not killing or stealing detail. It is just that the normal modes add in a lot of sharpened detail which make the image look like it is more detailed.

Gonna have to call you out on this one, too. ;-)

I didn't want to believe it, but for whatever reason, CINE is trashing low contrast detail, real detail, not artificial sharpening.

Compare stock CINE mode footage to Tv footage with -1 SHARPNESS. The high contast details will be virtually identical (little to no ringing in the Tv image) but the Tv footage will have preserved more low contrast detail, generally yielding less MPEG-2 artifacts.

I wish it weren't so, but it is.

[Eki Halkka]

Originally Posted by Thomas Smet

1. Cinema mode is softer which is the natural way it is supposed to be. It is not a reduction in detail but the lack of added electronic sharpness.
2. Normal modes are sharper because the DSP can work with the raw image to add sharpness to the raw image.

You choose which mode works best for you and which one gives the most pleasant look for you.

Exactly.

And i totally agree with everything else you wrote too ;-)