The Size of It All
I've been approached by countless fans of SciFi who protest the image scaling methods we use, simply because they don't understand the method.  Short attempts at explanation are sometimes successful with the more visionary individuals, but oftentimes people just simply don't get it.  After several years of this, I decided to create this section devoted to explaining the method, and instructing enthusiasts how to do it.

First, we have to understand what sources should be considered valid evidence for image scaling.  To conduct this kind of study, you must have a consistent set of rules in this area, or the work itself won't be consistent.  In Star Wars, the theatrical films are considered primary evidence (canon), with all other sources being secondary to that in varying degrees.  Star Trek adds televised episodes. Babylon 5 has, as the time of this writing, no officially published canon structure, so we at Babtech decided to establish a system similar to the two above to govern our work here.  Thus, we consider the televised episodes primary evidence, and all other sources secondary.  It is most logical to restrict ourselves to the actual show than it is to attempt to decide for ourselves what is or is not canon in other sources, where official word is less clear.  Whatever your particular project is, make sure you find out, or otherwise establish, these rules, and stick to it like glue.

Once the hierarchy of evidence is established, we must also understand how this evidence is to be used.  Fans who already understand the hierarchy of evidence still sometimes protest the actual measurement of onscreen images.  Some wonder how different camera lenses affect the relative scale relationship of objects onscreen.  Others protest that we don't know the relative distance between objects onscreen.  What these concerned fans lack is a full appreciation of the skill and intention of the filmmakers.

In the Lord of the Rings trilogy, the filmmakers used these tricks to create the onscreen images, like the one above.  If we were scaling the image above to determine how tall Gimli is, debating the methods used to achieve the effect is nonproductive.  This includes the relative distance between actors on set, the particular camera lenses used, image compositing, etc.  The fact that actor John Rhys-Davies is perhaps the tallest man present is irrelevant.  The concern is the height of the character Gimli in the finished image.  Other SciFi films and TV shows use similar methods to achieve certain visual effects.  For example, an Imperial Star Destroyer model would fit neatly in the back of my truck, where the finished image onscreen portrays a mighty mile long warship.  Ray Harryhausen used methods similar to the ones used in the Lord of the Rings films in many movies decades ago, like The 3 Worlds of Gulliver in 1960.  The advent of CGI simply makes this process faster and easier.  So, if in the image above, Gimli appears just under 5 feet tall (for example), that was the intent of the filmmakers all along.  It stands as canon evidence, trumping any other sources.

If we intended to establish an upper limit on Gimli's height, this could easily be done from the image above if we had reliable information about the height of Boromir.  Again, the size of the actor in question is irrelevant, as are camera lenses, image compositing, etc.  But if the height of the character is known within reason (possibly from scaling his height against an object of known size in a different image), Gimli cannot be beyond a certain height, because he is closer to the camera than Boromir.  Being closer to the camera, he appears slightly larger than he really is when compared to the more distant character.  We don't need to know how much closer he is, only that he is closer.  If Boromir appears X tall, and Gimli appears Y tall, Gimli cannot possibly be taller than Y.  If he was exactly the same distance from the camera as Boromir, he would appear ever so slightly less than Y.

This is the basis for scaling images in SciFi analysis.  We don't need to know the exact distance apart the characters are.  We are not attempting to establish an exact size on Gimli's height.  That is nearly impossible.  But we can easily establish that he cannot be taller than Y.  That is the nature of an "upper limit."

Suppose we also know how tall Pippin is.  We can clearly see that Pippin is closer to the camera than Gimli.  We don't know how much closer he is, and frankly don't care.  Again, we cannot establish an exact size on Gimli, so we won't even attempt.  But, knowing that he is more distant is enough to establish a "lower limit."  If Pippin appears A tall, and Gimli appears B tall, we know that Gimli is at least B tall.  Being more distant, Gimli appears slightly shorter in relation to Pippin than he really is.  How much isn't relevant.  We just need to know that if Gimli and Pippin were at the exact same distance from the camera, Gimli would appear slightly taller than B.

So, now we know exactly how tall Gimli is, right?  No.  And we won't from image scaling.  Is that relevant?  No.  We have established that Gimli is at least a certain amount shorter than Boromir, and a certain amount taller than Pippin.  That is all we can determine from the image, and that is all we intended to determine from the image.

Suppose (for example) Gimli appears slightly taller than 4 feet compared to Pippin.  Suppose (for example) he appears slightly shorter than 5 feet compared to Boromir.  We know only that he must be taller than 4 feet and shorter than 5 feet.  Those are lower and upper limits, respectively.  Using this method, we will never be able to determine exactly where Gimli's height measures between 4 and 5 feet, only that it is within that range, no matter what.

Suppose a fan insists that Gimli must be at least 5'5" tall, because...because...because he wants it that way?  He is simply incorrect.  Suppose someone writes a fanfic where Gimli is 2'3" tall?  It is simply incorrect.

Only sources that show Gimli between 4 and 5 feet tall can be taken seriously, because that is the way he looks in the canon film.  No amount of argument about the height of actor John Rhys-Davies, camera angles, lenses, relative distance, etc., is relevant to the finished film image.  In fact, those things were used to arrive at the finished film image, which is primary evidence, trumping anything else.

There have been analyses done by some who did not take into account the upper and lower limits, variable distance, etc., thinking that an actual size could be reached.  Those faulty methods usually lead to the assumption that the filmmakers were not careful to maintain proper scale.  While errors do occur, it is a rare and unfortunate event.

In case you still aren't convinced that filmmakers take great care in this area, there is a wonderful featurette on the bonus disc of Disney's Bambi: Platinum Edition DVD release that explains techniques used in this cartoon over 60 years ago.  The "multiplane camera" is actually mentioned twice on the disc, and was used to create a 3D background with a great sense of depth and scale.

In this image of a flat painting, zooming in on the house creates a false scale with the moon.  The moon is so distant that zooming (or walking) closer to the house should have no effect on the perceived size of the moon.  But zooming in on the house zooms the moon too!  Disney knew that this is incorrect, so they designed the "multiplane camera," where they paint different parts of the painting on different plates of glass.  These plates can be moved independently of each other, so you can zoom in on the house, but make the moon keep its distance!
To view a portion of this featurette, click here.  Highly recommended viewing, and it's only about 5MB.  Other filmmakers use tricks of the trade to maintain proper scale too, particularly with matte paintings, and more recently, CGI.  To hear an audio clip from the Predator Special Edition DVD describing on-location measurements, click here.  If you still aren't convinced, click here to see legendary Ray Harryhausen measuring models for his fairy tales.  It is a simple fact, and it is best to use it than to blindly deny it.

What if the height ranges from X to Y in the film, and the book/novel/comic says it's Z?  If Z is between X and Y, the prosecution rests.  If Z is not between X and Y, you have to figure out which source is primary evidence, and that one is correct without compromise.  You can't take this figure from this source, and that figure from that source at every whim.  That is poor methodology, and invariably leads to an inconsistent set of commentaries.  You decide which source is primary, and that one rules, period.

What if the size ranges from X to Y in the film, and (insert professional title) says otherwise?  If the film is your primary evidence, then this person is simply wrong.  There is a similar instance to the above in real life involving Babylon 5, where a concept designer scaled everything up after the show had been made.  These numbers were accepted by many, published on websites, etc.  But they have never been published officially, and aren't reconcilable with onscreen evidence.  The serious analyst can either accept those numbers, or accept evidence from the show, but never both.  An analyst who takes a little from here and a little from there is not to be taken seriously.  Once again, you establish the rules and stick to them, period.  Because there is no officially published hierarchy of evidence, we at Babtech use the show as primary evidence.  In our system, behind the scenes comments are irrelevant.  Even statements made by JMS himself are trumped by onscreen evidence.

But what if something JMS says makes the Minbari stronger than the Federation in Star Trek, so I can pound it into the heads of those Trekkies on the fanboards?  It doesn't fit into the hierarchy of evidence we use, thus can only be used as secondary, or supporting, evidence.  Serious analysts are more interested in finding out the answer than proving a pre established mindset.  In other words, I'd rather find out if the Minbari are as militarily strong as the Federation in Star Trek than to try to make my favorite the strongest, or my favorite ship the biggest.  A scientist never sets out to prove his point.  He sets out to arrive at a theory.

What if 15 years of published secondary continuity gets changed because some fan finds out that the primary material is different?  Shouldn't the single primary source be below 50 secondary sources?  No.  Primary is primary and secondary is secondary.  A lie is a lie, no matter how much it is repeated.  If a man cheats on his wife, no amount of lying will change his guilt.  By the same logic, a false statement is false, no matter how much it is repeated.  For example, Lucasfilm's canon policy regarding Star Wars has no allowance for multiple secondary sources to overcome the films.  The true analyst sticks close to primary evidence, using secondary sources only as supporting evidence to back up his findings.

Now that we've covered the basics of upper and lower limits, and following the hierarchy of evidence, we'll discuss the actual scaling techniques.  The important thing to remember in scaling work (or any other SciFi analysis) is to look for limits.  Find an object of known size, and find your target in a frame with that object.  In Babylon 5, we are told repeatedly that the station is 5 miles long.  But one episode gives us a more specific length.

This image quantifies the station's length down to half a meter.  It is exactly 8064.5 meters long.  We can then grab images for the broadside of the station, and establish the size of different areas of the cylinder.  The front "face," as we call it, of the station contains the docking bay, and is thus a valuable reference when scaling ships that dock with the station.  If we establish that this "face" is about 333 meters in diameter, and the bay opening is about 65x35 meters, we can determine the size of ships that enter the bay!  Easy, huh?

But how do we measure them?  They were created in the computer, thus there are no models to measure!  Just like in the Gimli example above, where the actor's size is irrelevant, so is the fact that we can't measure models.  The finished image is primary evidence.  Images on your TV screen or computer monitor are made up of tiny dots called "pixels," which is short for "picture element."  Put your face right up to the screen and take a look.  See those tiny dots?  Those are pixels.  It doesn't matter if you have a curved screen, flatscreen, bigscreen, plasma, or front projector.  You're still watching thousands or millions of little dots.  Those dots do the same relative thing from set to set, monitor to monitor.  That is what the TV signal tells your TV to do.  These dots are the smallest component of the TV picture, kind of like how atoms make up matter, but pixels aren't that small!  These dots can be counted.  Being the smallest building block of the picture, this provides the most precise available measurement of onscreen objects.  For example, it is more precise to measure your height in millimeters than in miles.

But counting pixels on your TV set is no good, because of several reasons:

Taking a picture of your TV screen is no good, because you might be at a very slight angle in any direction, distorting the picture!

So how do we measure the blasted thing?  I have a device that digitally captures the TV signal in realtime, saving it as video on my PC.  It reads the signal just like a TV, and stores the pixel instructions, recreating it on playback, within 5% of absolute perfection (more precise than nearly any standard TV set).  Computer monitors display slightly differently than TVs, so even if my monitor's geometry isn't perfect, the pixels will be perfect - even if the image is stretched a certain way, the number of pixels each way is unchanged.  If a character's head is round or stretched, it is still the same number of pixels in either case, and that number is what the TV signal said it should be!

So, then I hold my head up to the monitor?  No!  Then you load these images into an editor and let it count them digitally.


I took this photo with a digital camera moments ago.  Please excuse the clutter on my desk.  The program onscreen is VirtualVCR, which is my favorite video capture utility.  This capture was taken from The Weather Channel during a commercial break.  Yes, I can capture from cable, VHS, or even DVD.  The video plays in that window just like it does on any TV, and I can even make it fullscreen if I want.


After the capture, I load the video into my favorite video processing utility, VirtualDub.  I can clip out portions of the video I want to keep, prepare it for the internet, etc.  The left frame shows the raw video, the right shows the output after filtering (cropping, resize, etc.).  But what I was doing here was finding a specific frame to save as a still image.


I then load the still image into Paint Shop Pro, my favorite image editor.  PSP displays pixel numbers in the bottom left corner, just above the Windows "Start" button, indicated in red here.  In this case, the tip of the cursor arrow is on pixel number 305, 77.  The 305 is from left to right, and the 77 is from top to bottom.
PSP tells me exactly what pixel the cursor tip is on.  I can go to this side of an object and note the pixel, go to that side of the object and note the pixel, and subtract.

In this image, the cursor arrow is on pixel number 373, 80.  This means that in this image, the starfighter bay appears 68 pixels wide (373-305).

This gives an exact pixel count, in a few seconds, with minimal effort.  This works both up and down, but not at the same time!  If the object is angled (more often than not), I draw triangles on the object and apply Pythagoras.

But why all the precision and perfection if all we end up with is a range?  I can't control the fact that ranges are the result, but I can control how precisely I arrive at that range.  A precise measurement can not only make sure you're in the correct range, but also narrow down that range.

This process is repeated until you're satisfied that your images are all telling you similar things.

But each upper limit is just an upper limit from that image, right?  Yes.  That's why you do this over and over.  What if you come up with upper limits of 100, 300 and 1107?  No problem.  These are telling you the same thing.  The object can't be any more than 100 units, so it can't be any more than 300 or 1107 either!  If you did everything right, and particularly if you get similar results from different scenes, 100 is your best upper limit, because it is closer to the actual value.  The higher values are simply more generous upper limits.  The true value can't be higher than this.  It can be anything lower.

Similarly with lower limits, but the opposite.  The object can't be less than this value.  If you have lower limits of 30, 55, and 87, your best lower limit is 87, because it is closest to the actual value.  The others are simply more conservative lower limits.  It can be anything above this.

This is a simple illustration that shows the relationship of these values.  The upper limit represents the highest possible value.  The lower limit represents the lowest possible value.  The actual value is somewhere in between.

We don't always have both upper and lower limits, leaving a quite open-ended study.  For example, in my old Star Wars studies, I calculated lower limits on the firepower of turbolasers.  That tells us that these weapons are at least so powerful.  But we have no idea what the upper limit is.  It could be a little above the lower limit, or it could be a billion times higher.  But when both upper and lower limits can be established, and they are relatively close to each other, the study has been particularly successful.

How do you know when one object is closer than the next?  You have to look for overlapping.

For example, in this image, the upper fin of the Minbari Warcruiser is between the camera and Babylon 5, indicated by the red circle.  This tells us that the Warcruiser is closer, presenting an opportunity to calculate an upper limit on its size, because it looks bigger compared to B5 than it really is.
In this image, the Minbari Warcruiser is farther from the camera than the Omega Class Destroyer, because the Omega's antenna is between the camera and the Warcruiser's fin, indicated by the red circle.  This allows a lower limit to be calculated, because it looks smaller compared to the Omega than it really is.

Again, we can't establish the exact size of the Warcruiser in either case.  But frankly, we don't care.  That was never the intention.  We just want a range, from lower limit to upper limit, and anything in between is acceptable as the true size.

Okay, now how do we convert all this into a size range again?  The first thing you need to do is find an object of known size.

In this image, we can clearly see the pilot's helmet inside the cockpit.  If you look close enough, you can see his face.  You can measure the width of his face against the width and height of the cockpit.  It is safe to assume that the average human head is about 20 centimeters wide.  Otherwise, you can measure a similar helmet anytime a character is holding one onscreen.  So, now we know the approximate size of the cockpit.
Then we can scale the cockpit against the rest of the fighter.  Now we know the approximate dimensions of a Starfury, scaled against a Human head, with a margin of error that could be measured in centimeters.  We can then scale other onscreen objects, anywhere a Starfury is present.  But don't assume that you are correct within a few centimeters on the size of a huge warship!  Just know that you are close!

That's the basics of pixel scaling for SciFi analysis.  Now you can either experiment on your own, in which case I'm glad to help you on an individual basis; or simply understand the methods used better.

Back