Phase One IQ4 - Feature Update 1

[dougpeterson]

Thanks to DT for the 50 iso raws. After downloading and really going over the seashore shot, the results are both interesting and impressive.

The seagulls in flight are gone, as are the objects (foam cluster) on the water in the averaged shot. The color appears more pleasing to the eye also. As for noise, yes, the darker cliffs do show better in the frame averaged shot, however what catches me eye is the fact of how much more shadow gradient is available. The noise is less also, but the shadows show more color if that is possible. The water also takes on a more pleasing color but not sure that was something done in post. Adding a bit of sharpening with Topaz really pulls the darker cliffs out for more details, and no noise.

Shadow color quality, shadow noise, and shadow gradation are all sides of the same coin*; as you see in the raw, frame averaging improves the shadows across the board. Looking at the JPGs online the difference is meaningful but relatively subtle; looking at the raws the difference is really quite something.

Do note, while we appreciate your thanks, credit for those raws goes to the talented Paul Reiffer. I've done a lot of beta camera testing and it's very challenging to create a good test image 1) on the short timeline such things must be done on 2) demonstrates the technical attribute that is meant to be demonstrated 3) looks pretty.

*I guess that makes the coin three sided? My metaphors get worse as the day wears on.

 

[trond]

And the frame averaging raw files are now posted to our site (free to download). Go nuts!

Dear Doug,

Thanks for sharing the files!

Looking at the lighthouse in the distance, I see some strange artifacts/ghosting on edges of windows and cliffs.

It is apparent on both frame averaged and single shot images: DT16i_01, _02 and _05.

In general I see a slight loss of resolution in the frame averaged shosts, apparent in stationary objects like cliff texture and stones.

I would guess that tripod stability, wind and ground vibration are a real challenge in getting the most out of frame averaging!?.

It would be usefull to know the number of frames in each of the shots, and the topal elapsed time for each shot.

Best regards

Trond

 

[Paul2660]

Dear Doug,

Thanks for sharing the files!

Looking at the lighthouse in the distance, I see som strange artifacts/ghosting on edges of windows and cliffs.

It is apparent on both frame averaged and single shot images: DT16i_01, _02 and _05.

In general I see a slight loss of resolution in the frame averaged shosts, apparent in stationary objects like cliff texture and stones.

I would guess that tripod stability, wind and ground vibration are a real challenge in getting the most out of frame averaging!?.

It would be usefull to know the number of frames in each of the shots, and the topal elapsed time for each shot.

Best regards

Trond

Good catch, I missed the issue around the lighthouse. It's there as almost a shadow of the building and when you sharpen the image the artifact becomes much more evident. It's a movement type of artifact which is strange since I can't understand why the building would creat an issue like this. I don't think this is tripod movement as it would show up with the cliff also? The cliff has a slight bit of structural shift in the frame average, can only see this in a side by side comparison. But no artifacting. This type of issue would IMO really cause a problem with a city scape as all the buildings might pick up a similar artifact?

The slight loss of details I feel is the loss of noise, as if you zoom into the non frame averaged image there is some noise in the shadows, which the frame averaged image doesn't have.

Running Topaz Sharpen AI (stabilize) brings back anything lost for me but really brings the building haloing/artifact out and much more noticeable.

Paul C

 

[trond]

Good catch, ...

Paul C

Dear Paul,

What puzzles me here is that the ghosting is visible also in the single shot DT16i_02, as a slight "negative" copy of the window frames.

It looks almost like it remains from a previous capture.

The single frame is shot four minutes before the averaged shot, and still the ghost "survives" into the averaged frame!?

Even more strange is that the ghosting is apparently only in this area of the image.

Best regards

Trond

 

[Paul2660]

Hi Trond,

You have a good eye, I guess I used to have one. Yes it's there on the non frame average shot, albeit very faint. It's a movement type of issue to me, but not sure what would cause it, especially on the non frame averaged shot. Tripod movement? a tap? strange for sure.

But it happened on two totally different images taken separately, so I really can't figure that out. A C1 issue?

Paul C

 

[Anisometropia]

I found the same thing and discovered that the cause is a bad mask on the layer called "greenery". Turn it off and it goes away. Phew! How's that for an easy answer?!

 

[Paul2660]

Yes, it's a mask/layering issue, I had to turn off tip, greenery, and the blue enhancement mask to get it to get away.

Look at the "tip" layer with show mask on, the entire area is strangely effected, like a double exposure. Not sure what the tip layer is doing.

Paul C

 

[trond]

I found the same thing and discovered that the cause is a bad mask on the layer called "greenery". Turn it off and it goes away. Phew! How's that for an easy answer?!

Hm.. not sure about that.

The window "ghost" goes away if you turn off greenery, but the shadow of the lighthouse dome remains, and so does the cliff shadow "ghost".

Hopefully this is a C1 issue.

Best regards

Trond

 

[dougpeterson]

View attachment 142623

Yes, it's a mask/layering issue, I had to turn off tip, greenery, and the blue enhancement mask to get it to get away.

Look at the "tip" layer with show mask on, the entire area is strangely effected, like a double exposure. Not sure what the tip layer is doing.

Paul C

The two separate raw file captures are not in perfect registration, so if I copy-pasted a layer with a mask (after rasterizing the luma range) that would create the effect you're seeing. BUT... I don't see that on my copy, so maybe I uploaded an intermediate version? Checking...

UPDATE: O, I see... I only did final QC on the DT16i_03 and DT16i_04 raws because those are the images that we used in the article, but we ended up sharing/uploading several other raws including DT16i_01 and DT16i_02 on which I left shoddy masking. For now, I'll reset the adjustment of those images (and you can too... command+R) and rework them later today.

Note that the post processing was done by me, so blame us, not Paul Reiffer whose lovely image my poor masking made look weird. I thought it was important to show how the improved underlying shadow quality, especially in very deep areas of saturated subject matter (like moss and red stone) quality affected the ability to move this file deeply in post and still maintain a nice appearance.

Anyway, there’s no artifact in the raw; only an artifact in my shoddy post processing. Sorry for the confusion!

 

[Paul2660]

Thanks Doug.

No blaming just curious as to what happened.

There is enough in his file to see the overwhelming positive effects.

The issue is minor as I missed it totally even after sharpening.

Paul C

 

[dougpeterson]

The EIP raw files have been updated with the correct masks. We'll send a brief notice to anyone who previously downloaded them. Thanks for the heads up everyone!

I used the Refine Mask tool twice in a row, which is an amazing tool in C1 for this sort of thing (for dealing with high precision masking between images that are perfectly pixel-aligned to each other).

 

[Paul2660]

I keep seeing 1 to 1000 frames is the limit for P1 averaging.

So if I understand this correctly shutter actuation counts will sky rocket? Even though it all on ES?

Also, I am assuming the final shot recorded to the card is 1 raw not a layered raw that C1 has to work up, but not sure I understand the entire process.

Also something not mentioned is time.

Does a frame averaging shot forge effect of 7 minutes take a full 7 minutes, longer than 7 minutes or less? Obviously a single frame with ND would take 7 minutes.

Is the dark frame required which then keeps the camera out of use for the same time as previous exposure. I realize with the IQ4 with a single long exposure over 1 sec you can turn this off just curious if it comes back into play with averaging.

Paul C

 

[dougpeterson]

I keep seeing 1 to 1000 frames is the limit for P1 averaging.

In the time since I stated 1000 frames as the max, more work has been done. The max is now a bit more than 3000 frames.

So if I understand this correctly shutter actuation counts will sky rocket? Even though it all on ES?

In terms of wear and tear, there is no more wear and tear than a standard ES capture. On a tech camera this is zero wear and tear and on an XF the only wear and tear is opening the shutter and raising the mirror that happen when you are starting a sequence. In neither case with the duration of the frame averaging cause wear and tear.

That said, I’ve not checked how they record this in terms of shot count on the back. I strongly suspect they record it as one shot, since that’s what it is just about any way you think about it. But I will check tomorrow.

Also, I am assuming the final shot recorded to the card is 1 raw not a layered raw that C1 has to work up, but not sure I understand the entire process.

Correct. The process generates a single raw file just like a standard capture. No special post processing required.

Push button > get raw. Couldn’t be simpler.

Does a frame averaging shot for the effect of 7 minutes take a full 7 minutes, longer than 7 minutes or less? Obviously a single frame with ND would take 7 minutes.

If I understand your question correctly, then yes. If you want 7 minutes worth of blur (eg to turn a waterfall into silk) then you need 7 minutes of real world time. Time, as usual, is the only thing you can’t buy more of.

If on the other hand you are doing it only to decrease noise then you could do, say, a 4 second frame averaging of 1/4th of a second exposure for a total of 16 captures (under the hood) that has ~4 times less noise in the shadows than a normal 1/4th of a second exposure. Obviously the subject matter and tripod would need to allow for a 4 second exposure worth of movement.

Is the dark frame required which then keeps the camera out of use for the same time as previous exposure. I realize with the IQ4 with a single long exposure over 1 sec you can turn this off just curious if it comes back into play with averaging.

For a given frame averaging sequence only a single dark frame is required, and is the same length as the individual exposures, not the total length of the sequence. So if doing a 100 second frame averaging using 400 samples of 1/4th second, then a single dark frame will be captured at the end that lasts 1/4 second.

 

[Paul2660]

Doug,

Great info and really appreciate your assistance.

Looking forward to the firmware.

Paul C

 

[Boinger]

I keep seeing 1 to 1000 frames is the limit for P1 averaging.

So if I understand this correctly shutter actuation counts will sky rocket? Even though it all on ES?

Paul C

So there is basically no actuation with an Electronic shutter it's a misnomer. When you take the picture you freeze the latest capture of the shutter but it is not actually doing anything different than live view.

When you do live you with your back it is essentially doing 1/60th actuations repeatedly until you chose to stop live view or take the picture.

 

[Paul2660]

Still need to discover the effective use of this tool with slight subject movement.

In most of my landscape work there will be slight wind moving tree limbs and leaves. Curious as to how these will be effected.

With focus stacking movement even slight is a problem. So I really don’t use it much. Both of the stacking software tools were not able to handle slight movements and this created voids in the final output.

The seagulls in flight are gone in the example and I guess so would cars and people. Curious on other natural subject matter.

Paul C

 

[alistairsimmons]

Still need to discover the effective use of this tool with slight subject movement.

In most of my landscape work there will be slight wind moving tree limbs and leaves. Curious as to how these will be effected.

With focus stacking movement even slight is a problem. So I really don’t use it much. Both of the stacking software tools were not able to handle slight movements and this created voids in the final output.

The seagulls in flight are gone in the example and I guess so would cars and people. Curious on other natural subject matter.

Paul C

Hi Paul

I have an image shot this morning with the BETA Firmware, with slight movement in trees if you want to email me.

A

 

[trond]

Quote Originally Posted by Paul2660:

Still need to discover the effective use of this tool with slight subject movement.

In most of my landscape work there will be slight wind moving tree limbs and leaves. Curious as to how these will be effected.

With focus stacking movement even slight is a problem. So I really don’t use it much. Both of the stacking software tools were not able to handle slight movements and this created voids in the final output.

The seagulls in flight are gone in the example and I guess so would cars and people. Curious on other natural subject matter.

Paul C

Dear Paul,

If you poke around in the image at 200%, you can see faint shadows of people moving around.

See the screenshot above at the top of the hill at the end of the walk way.

Best regards

Trond

 

[trond]

Here is one more example with people more visible.

If shot with a ND10, these people would disappear into a faint foggy blur.

Best regards

Trond

 

[dougpeterson]

View attachment 142642

If you poke around in the image at 200%, you can see faint shadows of people moving around.

See the screenshot above at the top of the hill at the end of the walk way.

Best regards

Yes, but this particular frame averaging was also done at 1/125th shutter speed free frame. I suspect they would have ghosted more thoroughly if either a slower shutter speed (e.g. 1/8th) was used per frame, or if more total frames were averages (or both).

If used at a shutter speed that is gapless the effect should be identical to a strong ND filter but with lower noise. That is, up to the longest supported total capture time at which point the strong nd filter would win out again.