Sharpness and AA filter on DSLR?

bringing up 5D2 footage to this discussion makes no sense - yes, it looks sharper, that's the whole "false detail" thing we all know about

the question here is whether the OLPF hack brings improvements on any of the two main aspects of "sharpness":
* resolution: ability to show finer detail
* acutance: sharpness of edge transitions
and what downsides it may have (like increased aliasing/moire)

http://www.lensrentals.com/blog/2009/06/have-you-seen-my-acutance


edit:
ok, I delved into that very long thread...
I understand this is the part of that thread where you say "the OLPF myth was debunked"
http://www.dvxuser.com/V6/showthrea...ved-olpf-removed-for-higher-resolution/page47

it indeed seems to add no resolution at all, but the real-world samples definitely *look* a lot sharper, so I'd say it brings an improvement acutance - that's precisely what "sharpening in post" can do too, so I definitely wouldn't open the camera to get rid of the OLPF filter - yet, if magic lantern can disable it, images would be sharp directly out of the camera, without all those IQ issues that arrive because other stuff goes when the filter goes (IR filter, etc)
 
Last edited:
Samuel H said:
bringing up 5D2 footage to this discussion makes no sense - yes, it looks sharper, that's the whole "false detail" thing we all know about
<SNIP>
it indeed seems to add no resolution at all, but the real-world samples definitely *look* a lot sharper, so I'd say it brings an improvement acutance - that's precisely what "sharpening in post" can do too, so I definitely wouldn't open the camera to get rid of the OLPF filter - yet, if magic lantern can disable it, images would be sharp directly out of the camera, without all those IQ issues that arrive because other stuff goes when the filter goes (IR filter, etc)
Click to expand...
The point is that the clip I posted looks super sharp too. Just because the couple of clips from the cameras with the OLPF removed look sharp doesn't mean that removing the OLPF actually improved anything.

It's over a month since this nonsense started. Don't you think that if this really were some major breakthrough in improving 5D3 video that the guys who hacked their cameras would have been trumpeting it all over & posting more video 'proof'? I suspect that those who damaged their cameras by removing the OLPF are either keeping quiet because they are embarrassed to have been fooled or desperately trying to fit the OLPF back into the camera.

The OLPF is not an electronic component it is piece of glass there is no way that Magic Lantern firmware can affect it.
 
usually the OLPF is not electronic, but this one seems to have too much effect on videos, and none in stills, so maybe it is electronically controlled (e.g. its effect is bigger when a current is running through it), and it may be possible to control it from ML

or maybe it's a standard OLPF and people haven't noticed the improved acutance in stills because they were looking at resolution

we'll see
 
nigelbb- a 720p video is not sufficient to judge sharpness from a 1080p source video. My 5D2 worked fine most of the time (still have it), however when significant aliasing appears, the shot is ruined. Post sharpened 5D3 1080p footage is significantly more detailed than 5D2 footage, especially in vertical resolution (from test chart comparisons).

The OLPF is an electronic component: it's also the dust vibration system. Pulling the plug affects a circuit which could change firmware behavior. If that is happening, then a firmware change could have the same effect. Note that Bill's MaxMax mod had the vibration connector plugged in and he did not get the results James Miller obtained with it completely removed.

If James Miller's beach scene can be matched with a stock 5D3, we'd love to know exactly what lens and settings he used so we can try to repro. I can (almost) match his first city clip with post sharpening. Straight out of the 5D3, footage is very soft (I have two 5D3's- they both behave the same). Do you have a 5D3? If so, perhaps you can post clips showing it matching James Miller's beach or town shot in sharpness and let us know how you did it (if you upload the raw file to vimeo we can be sure it was not post sharpened).
 
My point in posting the 5D2 clip (which is full 1080p BTW & you can download the raw .MOV file) is to show that a single video is no proof of anything. I use that clip because through some freak trick of the light it really stands out & it was one of the very first clips that I took over 3 years ago when I got the 5D2 so it's the original firmware & auto-everything. I could offer that clip as proof that the original firmware & auto-everything delivers a sharper image than later manual firmware but that would be nonsense. Just as it's nonsense to show a clip recorded with the OLPF removed that looks sharp & say that is proof that the hack improves resolution.

The OLPF is not an electronic component it's a glass sandwich. The dust vibration device is glued to it.

The biggest confirmation for me that removing the OLPF does not do what is claimed is that despite over a month passing since this was first discussed that we have seen no further 'proof'. If it was indeed such a miracle cure then James Miller & Andrew Reid would be writing up blog posts & tutorials & posting videos to show how everyone could do it.
 
nigelbb said:
The OLPF is not an electronic component it's a glass sandwich. The dust vibration device is glued to it.

The biggest confirmation for me that removing the OLPF does not do what is claimed is that despite over a month passing since this was first discussed that we have seen no further 'proof'. If it was indeed such a miracle cure then James Miller & Andrew Reid would be writing up blog posts & tutorials & posting videos to show how everyone could do it.
Click to expand...

They've posted at least three videos. They didn't remove the OLPF optical component by itself, they also removed the dust vibration electronics. Their "OLPF" removed videos included the electronic components as well.

Andrew Reid's site is ad supported. His posts are designed to generate traffic. It could be that he's waiting for traffic to die down before releasing more material.

If you do not own a 5D3, I'm not sure I understand the point of your posts? As a 5D3 owner, I'd like to know how James Miller was able to achieve such a high level of apparent sharpness without post sharpening. What lens, f-stop, ISO, etc. In my tests I discovered that a variable ND increased apparent sharpness (actually acutance) in bright light (likely due to the circular polarizer, which is known to improve microcontrast). It did not increase resolution on a chart. James Miller also used a variable ND for the beach video. I think there's still more to be learned here.
 
I am not sure why you seem so convinced that the dust vibrating mechanism has some effect on the image. It's for shaking the dust off nothing more.

I agree that Andrew Reid has already moved on to new enthusiasms like the Black Magic Cinema Camera but honestly if this OLPF removal were the magic fix that he was claiming then he would be crowing about how great it was. He now just ignores the subject.

My wife & I each own a 5D3 & a 5D2. I would be really pleased if there was a shred of evidence for this OLPF removal malarkey but there isn't & it has now reached the level of an urban myth that needs to be debunked.

I'd like to know how I achieved that high level of apparent sharpness with the 5D2 on auto-everything. I was using the 24-105mm F/4 IS lens but unfortunately I have lost the THM file so don't have the EXIF info regarding aperture, shutter etc. but there was no ND or polarising filter. It was shot at a quayside cafe in Golfe-Juan on the French Riviera at 08:37 on 25th April 2009 & I think it looks so good because it was very bright sunlight that was at a low angle as it was early in the day which gave some hard shadows & good contrast. These are the conditions that make the 5D2 look good & the second seaside shots from James Miller are also in very bright sunshine.
 
nigelbb- I only bring up the dust system as it is an electrical component. As a software/hardware guy I understand that any time an electrical component is changed, it's possible the software (firmware) can detect said change, and adjust software behavior. The OLPF is designed to (effectively) remove frequencies higher than the Bayer sensor array spacing (so the frequencies which map to greater than 1/2 the sensor resolution are attenuated or cut). That's why so many folks are puzzled that removing the OLPF has a significant effect on video. One likely explanation is that since the photosites are binned (averaged together) for video, that performs an effective low-pass filter and explains why the OLPF removal does not show significant aliasing. In computer graphics imaging we call that super- or multi-sample anti-aliasing.

Compare James Millers' bricks & tiles/shingles in this footage:
https://vimeo.com/39541733

to this footage (shot with one of the sharpest Canon L lenses (70-200 F2.8L II) in very bright light:
https://vimeo.com/39552988

You can download full 1080p originals from vimeo for both videos.

I had to significantly sharpen the brick video. My understanding is that James Miller did not sharpen his video (even if in-camera was on slightly, it doesn't do much). Study the shots of the tall pointed building's shingles. Sharp and detailed, but no aliasing. You can try to repro yourself with your 5D3 by finding a similar object at a similar distance (I shot bricks a varying distance/zooms to explore as much as possible). If removing the OLPF & dust system doesn't do anything, how did James Miller achieve such detail and sharpness without software sharpening?
 
From my point of view, the street scene looks way more detailed than usual 5D footage. Whatever is going on is an improvement to the detail level in the final image. As to comparing it to the 720p table footage, they are two different shots. The 5D's weak point is detail in wide shots. Medium to closeups are much less of a problem. If you had a 5DMKII next to the MKIII in that street scene the MKII would not have had as much detail.

This is worth keeping an eye on as detail is the only thing holding the 5DMKIII back from being "the camera" in many ways.
 
gene_can_sing said:
I bought into the whole DSLR revolution in 2008 and bought a ton of Canon lenses thinking they would fix all the known issues and we would have a great camera. But year after year, it was just one disappointment after another. The latest 5D3 was just such a let down in a series of let downs.
Click to expand...

Great camera designs, rationed out and delivered by corporate accounting strategists. Good enough to get fantastic world class film like images, but not all you dreamed for. Bitter sweet this DSLR revolution for some, but just sweet for others.
I Just remember not long ago, buying a 400 foot can of Kodak 35mm film, processing, daily and transferring it scene by scene to HD tape via Spirit and get only 4 minutes of raw footage.
Today for the price of that, I can add one new Canon L lens for about every can of film I used to shoot. This is sweet.
 
Last edited:
You must log in or register to reply here.
Back
Top