Starship Asterisk*

Chris Peterson wrote:

geckzilla wrote:I saw this attempt at using morphing software to interpolate data between frames of a rotating view of comet 67P. I think it's a very good example showing what kind of false movement morphing software can give to things. This is much easier to understand since we know the comet does not bend and flex like this.
https://www.flickr.com/photos/105035663 ... otostream/

Ouch.

The big difference here is scale. The camera is close to the comet and there is a lot of relative rotation. But the camera is 20,000 light years from V838 Mon.

It doesn't matter where the camera is. It's an example of how morphing software interpolates data. It doesn't understand what's between so it does the best that it can.

geckzilla wrote:It doesn't matter where the camera is. It's an example of how morphing software interpolates data. It doesn't understand what's between so it does the best that it can.

Well, some data are more suited to interpolation than others. The great distance of V838 Mon means that interpolation between the sequence of recorded images, is likely to be more reliable than what we see done with the comet clip. Many more critical things are changing between the recorded comet images.

Nitpicker wrote:

geckzilla wrote:It doesn't matter where the camera is. It's an example of how morphing software interpolates data. It doesn't understand what's between so it does the best that it can.

Well, some data are more suited to interpolation than others. The great distance of V838 Mon means that interpolation between the sequence of recorded images, is likely to be more reliable than what we see done with the comet clip. Many more critical things are changing between the recorded comet images.

I disagree. The clouds having this elastic property is just less likely to trigger in your mind a "this is wrong" sense. The movement of the clouds is just as egregious though it takes much more careful study to notice because your brain easily accepts the morphing of the clouds.

geckzilla wrote:

Nitpicker wrote:

geckzilla wrote:It doesn't matter where the camera is. It's an example of how morphing software interpolates data. It doesn't understand what's between so it does the best that it can.

Well, some data are more suited to interpolation than others. The great distance of V838 Mon means that interpolation between the sequence of recorded images, is likely to be more reliable than what we see done with the comet clip. Many more critical things are changing between the recorded comet images.

I disagree. The clouds having this elastic property is just less likely to trigger in your mind a "this is wrong" sense. The movement of the clouds is just as egregious though it takes much more careful study to notice because your brain easily accepts the morphing of the clouds.

Thing is, I don't see the motion of the clouds at all. I see the motion of the light echo through the (comparatively stationary) clouds. And I see the light echo moving through the clouds more smoothly when it is interpolated. Not to say it is entirely true or even necessary. Without more recorded data we can't actually determine how good or bad the interpolation is in this case. But in the case of the comet, one can immediately see that that data were not well suited to interpolation.

Nitpicker wrote:Thing is, I don't see the motion of the clouds at all. I see the motion of the light echo through the (comparatively stationary) clouds.

That's not what I see. I see as bad a morphing failure as with the comet. And that's not surprising, given how morphing software is designed. It interpolates motion. When there is no motion, artifacts like we see in the light echo video are very likely.

Chris Peterson wrote:

Nitpicker wrote:Thing is, I don't see the motion of the clouds at all. I see the motion of the light echo through the (comparatively stationary) clouds. And I see the light echo moving through the clouds more smoothly when it is interpolated. Not to say it is entirely true or even necessary. Without more recorded data we can't actually determine how good or bad the interpolation is in this case. But in the case of the comet, one can immediately see that that data were not well suited to interpolation.

That's not what I see. I see as bad a morphing failure as with the comet. And that's not surprising, given how morphing software is designed. It interpolates motion. When there is no motion, artifacts like we see in the light echo video are very likely.

What makes you think that extra recorded images between the current recorded ones, would look critically different from the interpolated images? The illusion of cloud motion, as the light echo moves through it with its unique ellipsoidal shape, might be quite realistic.

Nitpicker wrote:What makes you think that extra recorded images between the current recorded ones, would look critically different from the interpolated images? The illusion of cloud motion, as the light echo moves through it with its unique ellipsoidal shape, might be quite realistic.

Because in this example, I see actual structure being replicated, shifted, and blended in the interpolated frames. It's simply wrong. A light echo can't look like that. There are better light echo animations out there that only use actual images.

Chris Peterson wrote:

Nitpicker wrote:What makes you think that extra recorded images between the current recorded ones, would look critically different from the interpolated images? The illusion of cloud motion, as the light echo moves through it with its unique ellipsoidal shape, might be quite realistic.

Because in this example, I see actual structure being replicated, shifted, and blended in the interpolated frames. It's simply wrong. A light echo can't look like that. There are better light echo animations out there that only use actual images.

It is exactly what an interpolation is and must be. It is an approximation of the missing data, based on the real data. You haven't shown it to be "simply wrong", you just don't like the results of the approximation. I find your strong opinion on this matter rather odd. My more moderate opinion is that I don't think the interpolation is terribly necessary, but I don't think it distracts or detracts, either.

Nitpicker wrote:It is exactly what an interpolation is and must be. It is an approximation of the missing data, based on the real data. You haven't shown it to be "simply wrong", you just don't like the results of the approximation. I find your strong opinion on this matter rather odd. My more moderate opinion is that I don't think the interpolation is terribly necessary, but I don't think it distracts or detracts, either.

My opinion is strong because it isn't an opinion. It is fact based. The image misrepresents reality. Blatantly. I do know, with 100% certainty, that this is not what a light echo looks like, nor could it.

Interpolation could be used to reasonably fill in the missing frames. But morphing software can't perform the necessary interpolation, because it is based on the underlying premise that material is moving between frames. You'd need an entirely different algorithm, which essentially propagated a wave of low frequency luminance data over a set of frames that didn't allow the high frequency information (physical structure) to move.

Chris Peterson wrote:

Nitpicker wrote:It is exactly what an interpolation is and must be. It is an approximation of the missing data, based on the real data. You haven't shown it to be "simply wrong", you just don't like the results of the approximation. I find your strong opinion on this matter rather odd. My more moderate opinion is that I don't think the interpolation is terribly necessary, but I don't think it distracts or detracts, either.

My opinion is strong because it isn't an opinion. It is fact based. The image misrepresents reality. Blatantly. I do know, with 100% certainty, that this is not what a light echo looks like, nor could it.

Interpolation could be used to reasonably fill in the missing frames. But morphing software can't perform the necessary interpolation, because it is based on the underlying premise that material is moving between frames. You'd need an entirely different algorithm, which essentially propagated a wave of low frequency luminance data over a set of frames that didn't allow the high frequency information (physical structure) to move.

Such a "fade in/out" algorithm would make no attempt to estimate what the structure looked like in the unilluminated space between the real frames. Morphing does, and whilst it can give the appearance of material movement, one only needs to know that the material isn't actually moving (comparatively) and that you are looking at ellipsoidal shells of light, moving through and illuminating the material, to not be distracted by the morphing. One always needs to understand a few basics to appeciate any astro imagery, otherwise it tends more toward abstract art. I'll concede that if one didn't know the material is stationary, it might be misleading.

Nitpicker wrote:Such a "fade in/out" algorithm would make no attempt to estimate what the structure looked like in the unilluminated space between the real frames. Morphing does, and whilst it can give the appearance of material movement, one only needs to know that the material isn't actually moving (comparatively) and that you are looking at ellipsoidal shells of light, moving through and illuminating the material, to not be distracted by the morphing.

I disagree completely. A model that considered what's actually happening physically could stand some chance of at least approximating reality. This video flat out misrepresents it.

Chris Peterson wrote:

Nitpicker wrote:Such a "fade in/out" algorithm would make no attempt to estimate what the structure looked like in the unilluminated space between the real frames. Morphing does, and whilst it can give the appearance of material movement, one only needs to know that the material isn't actually moving (comparatively) and that you are looking at ellipsoidal shells of light, moving through and illuminating the material, to not be distracted by the morphing. One always needs to understand a few basics to appeciate any astro imagery, otherwise it tends more toward abstract art. I'll concede that if one didn't know the material is stationary, it might be misleading.

I disagree completely. A model that considered what's actually happening physically could stand some chance of at least approximating reality. This video flat out misrepresents it.

I disagree with you, but not so completely.

Well, I thought the video I linked would lead to further understanding and agreement but it seems only to have increased disagreement and misunderstanding.

geckzilla wrote:Well, I thought the video I linked would lead to further understanding and agreement but it seems only to have increased disagreement and misunderstanding.

Well, it shows what happens when morphing goes wrong. But I don't think it helped because the situation was different. In principle, morphing software, properly applied, would work with that set of keyframes. The failure was of the automation (or the person managing it) to properly define matching reference points. There's no way that conventional morphing software will ever work on a light echo image, however, because you can't just pick up information from one region and blend it into another. With a light echo, nothing is moving. The finer structure is always left behind, although its relative intensity changes. That's just not what morphing software does.

(And your link is broken... I guess the author realized it needed some work.)

Ah, so it is. I didn't think it was bad work, personally. The rotation sequences are not actually frame by frame animations of the comet nucleus as Rosetta orbits. They are once-per-orbit mosaics and while at first look they appear continuous they are actually not! Each frame is composed of four separate images which must first be stitched together and for each of those frames the probe is at a slightly different position from the previous one. It's a lot of work and Thomas does it well. There are some more videos still in his Flickr gallery but he also posts them here in the submissions thread.

Chris, I think your fade in/out algorithm would be superior in cases where the whole cloud was already thoroughly illuminated to some degree, with a noticeably brighter light echo pulsing through it. Also in cases of a sufficientlty short time interval between real frames, where there might be little or no spatial gap between the leading edge of one real echo shell and the trailing edge of the next. But neither of those cases are in play in this APOD. I think the morphing technique is the better interpolation method in this case, but it would probably be just as good without any interpolation.

Isn't the answer that the material around the star is moving outward, just not as fast as the light echo?

Val

Guest wrote:Isn't the answer that the material around the star is moving outward, just not as fast as the light echo?

The material isn't moving enough to result in a significant change in position over the time range we've been observing this object.

I think the average person is left with the impression the dust is moving and that's why the morphing interpolation is egregiously wrong.

geckzilla wrote:I think the average person is left with the impression the dust is moving and that's why the morphing interpolation is egregiously wrong.

I dunno whether you can blame the morphing for that entirely. Just looking at an animated sequence of real frames might be enough to give the average person the impression that the dust cloud is moving. I think it is just the illusory nature of this particular light echo, which seems to be the only thing illuminating this dust and only in a thin ellipsoidal shell at any one time.

I will concede that there are parts which appear to be moving but there are also parts which are easily noticed as being quite stable when flicking back and forth between individual frames of real imagery. That stability and the sense that it is the light which is expanding and not the dust is lost completely with the morphing interpolation applied because it leaves none of them intact. With close inspection of a non-morphed sequence, one can deduce what is actually being witnessed. With the video, it is impossible. Telling people that the dust is not moving and then showing them this video will make them call you a liar. It's really confusing and that's not good at all for helping laypeople understand science. It's very good at causing intelligent people to have unnecessarily long and pointless arguments, too.

Left to our own devices, I very much doubt that the average observer -- myself included -- would have realized (much less deduced) that an animation of just the real frames was of a light echo, and not a physically expanding cloud. Thankfully, the truth of the matter was nicely explained in the caption of the APOD. That interpolations were applied in between the real frames, was also explained. So, after all that, if the average observer was still unable to ditch the notion that the cloud was physically expanding, that's their own silly fault.

Even after reading that it's smoothed it's very difficult. And it's not their fault. Not everyone has the time to put forth the effort required to find the original frames and scrutinize them. This particular object has had videos made about it several times now with new frames added on the end as they are acquired. The first time I saw it was before I had really gotten into astro imaging. I clearly remember staring at the video, reading the description of what a light echo was, re-reading it, staring at the video again, and being unable to reconcile the fact that the clouds are not moving.

It could be done better. It could be made more clear but it wasn't. This is the version that looks "good" despite its flaws. The fact that you continue to apparently defend these flaws is mind-boggling and frustrating to me. I can't tell if you yourself don't totally understand it or if you really think it's ok the way it is even though you know.

Show me an interpolated version of this particular object which looks better.

My criticism is in response to the savaging this APOD received from Chris and then you. I felt that your criticisms were wildly overstated. From one real frame to the next, there are so many parts of the dust cloud which show matching similarities in shape, but which are noticeably displaced in the plane of the image. The decision to interpolate may be questionable, but once made, I can't imagine a better interpolation method for this particular case, than the morphing method employed. Sure, the morphing does some strange things to the finest "structural" details of the cloud, but in terms of the bigger picture of being able to map the variable density of the dust cloud in 3-D, I don't imagine this approach is terribly flawed. Learning more about light echoes was one of the great aspects of this APOD. But learning of a light echo's ability to provide data to build 3-D spatial models, was even more interesting.