Moonlit Rail ™

The "Massachusetts Central Railroad"

A Very Personal Introduction

When I was a young chap of ten or so, I often visited my good friend, John "Urbie" Kafalas, at his home in Wayland, MA, by taking the commuter train out of North Station in Boston. This train followed the route of the Fitchburg RR to the Lincoln stop, one town north. Every time I asked why I could not take the train directly to the Wayland train station just down the street from where John lived, I received a wavering answer; this would bother me repeatedly.

Years later, as I was spending many enjoyable miles exploring "back roads" west of Boston by motorcycle, I would stop in for a late lunch at the general store located in the center of Petersham, MA. In the gift area located towards the rear of the store were a variety of books, many of them written by local authors about local history. One such book that caught my eye was author J. R. Green's The Mass Central: The Quabbin's Phantom Railroad — a story about a railroad which built most of a route through the Swift River valley before abandoning it in favor of an alternate — the very railroad whose station in Wayland MA I never had the "pleasure" of embarking or alighting from so many years earlier.

I bought that book right there. The more I read it, the more I realized just how many of the places I loved to explore were in fact connected by segments of that railroad, whether actually built as "railroad" or merely as curious movements of earth. That I could theoretically walk from my friend's house all the way to the Quabbin upon a now-abandoned railbed snaking through hills and forests in a corridor of nearly-undisturbed natural beauty was the most unexpected yet invigoratingly delightful thought! This was soon tempered with the sad realization that — like mold upon a fruit beyond its time — the "progress" of humanity would soon come to obliterate this remnant surviving on borrowed time.

To explore this wonder before it was gone, I would have to work quickly! I would have to photograph those portions of the route that were still undisturbed. I would have to map the route so as to reconstruct it within the confines of a digital simulation yet to be envisioned. I would have to out-flank the rails-to-trails scourge and its NIMBY zealots. The race was on!

Static Imagery

If you happen to fancy the Massachusetts Central themed background image used by default on this web site, you can find several versions for download on the railroad area page.


It has long been my goal to attempt to capture some of the flavor of what little remains of the Massachusetts Central in movie sequences, shot as if from the front of one of the engines traveling along the original route. Opportunities are fast dwindling, however, as a mold of asphalt grows upon remaining stretches of right-of-way, dotted with trash barrels, park benches, sign posts, inline-skaters, bicyclers and all manner of things anathema to a locomotive on a track.

What is left to capture on video is typically overgrown, making traversal with a video recorder impractical — particularly at any sort of speed approaching that of the original locomotives. And to do that would require a vehicle capable of realistic speeds; the closest facsimile — a "quad-runner" or consumer "all-terrain-vehicle" — would result in a very bumpy ride. At much greater expense, one could employ techniques long used in the motion picture industry, namely a camera mounted on a dolly rolling on temporary tubular track; the movie industry has another trick up its sleeve as well, the use of stabilized, hand-held cameras.

Of the various techniques to smooth out hand-held motion pictures, perhaps the best known is the Steadicam® invented by Garrett Brown. Employing heavy weights, joints with multiple axes of motion, and a harness to carry it all, the Steadicam allows a camera operator to move over uneven terrain without causing undo jarring of the camera shooting the footage. George Lucas went a step further in shooting Return of the Jedi by shooting with a Steadicam in slow-motion on a forest floor and then speeding up the resulting footage in post-production. This is just the sort of technique that could be employed in the making of my Massachusetts Central footage were it not for a few inconvenient truths — the cost of procuring a Steadicam, and then the darned overgrowth. (Subjectively, even sequences shot with a Steadicam in slow motion and then sped up the appropriate amount still look sped up, unnatural in some way.)

To solve this dilemma, I utilized a technique I haven't seen employed elsewhere: shoot a motion picture as a sequence of disconnected still images, and then hand-assemble them into a "movie" in "post production". As it turns out, this is much easier said than done. Actually shooting the individual frames of such a movie is the easy part; assembling them into a "movie" is some variant of madness. I'll mention how, below; but first, let's present the fruits of the first such effort.

Approaching the Tunnel

To test my hypothesis about the viability of making a motion picture out of a bunch of still frames, in May of 2001 I shot 1,001 "frames" of the track bed as it approaches and proceeds through the well-known 1,110 foot tunnel built to relocate the railroad around the Wachusett Reservoir. Perhaps due to the "madness" alluded to above, it took more than 9 years to align them all. And here we are!

The video in the object, above, is small (352x240) and runs at an industry-standard 29.97 frames per second — absurdly fast, as it equates to a ground speed of about 86 miles per hour. If the browser's video player doesn't work for you, or you'd like to download a more conventional MPEG-1 format video (which should work in just about every sort of video player), you can open this link in your video player; it's a bigger download with twice the dimensions, and runs at 12 frames per second — the correct speed for this track segment at a wee bit under 35mph.

As for the "madness" that delayed completion of the tunnel-approach video segment, a description of the process and its pitfalls is due. For sure, several lessons were learned in the process of assembling this movie clip, a few of which ought not be repeated next time.

Assembling Frames of a Motion Picture

The camera used to collect the still frames in the proof-of-concept tunnel-approach clip was a Casio QV-2000UX with a native resolution of 1,600x1,200 pixels — fairly common in the early history of consumer-grade digital cameras. Owing to the large number of pictures required and the diminutive size of the then-prevalent MPEG-1 video standard (352x288), I shot the frames at 800x600. (Evidently, I surmised correctly that by the time the footage was ready, common video formats would support 640x480 or above.) The larger frame size would allow me to shift each frame up/down/left/right so as to align it properly against its predecessor, then crop to the target frame size. If needed, I could also rotate each frame slightly while aligning.

I hadn't realized when shooting the frames just how much camera rotation changes each time the shutter is pressed. Most of the rotational corrections from the horizontal are less than 3 degrees; but most frames require at least half a degree of correction, which takes about as much time to execute in the image editor of your choice (I used the GIMP) as a much more severe angle of correction. And I learned the "hard way" that one should program one's digital camera for a fixed focus distance, fixed white balance, and in all likelihood, a fixed exposure. Having to correct for white balance and exposure for every frame slowed down the process to about 5 minutes per frame.

Here is the sequence of events employed to post-process each frame, which you can employ yourself should you embark on such an adventure:

  1. Open the next frame in your sequence, preferably one in a lossless image format (such as PNG or TIF) so that your changes will not introduce distortions.
  2. Paste the previous frame (which you saved on your clipboard) as a "new layer".
  3. Add an opaque third new layer which is your background color (black). Arrange the layers so that your new frame is at the top, and the background at the bottom.
  4. If you didn't set your white-balance and image exposure to fixed values when you shot your frames, now is the time to use the color tool "auto level" to (hopefully) normalize the tonality to match the other frames in your sequence.
  5. If your camera's image sensor has any stuck pixels in known locations, you can use the erase tool to remove them. (Five problem pixels needed to be erased from the prototype sequence above, most of them showing up in the dark tunnel interior.)
  6. Set the opacity of the top layer to 50% so that you can see the layer below (your previous frame) simultaneously.
  7. Use the move-layer tool to shift your semi-transparent top layer so that the "horizon" point lines up perfectly with that of the previous frame in the layer below.
  8. Set the opacity of your top layer back to 100% (opaque) and then click on the visibility check-box for your top layer repeatedly, toggling it on and off, so you can flip back and forth one frame in your movie sequence. This will likely show you if you need to move your top layer up/down/left/right a pixel or two to make the inter-frame alignment to your liking. This process will also likely show you if the two frames are out of rotation with respect to one another.
  9. If a rotational correction is required, now is the time to do it. Use the layer-visibility toggle to check how good your guess was as to the number of degrees of rotation; use your image editor's "undo" feature and re-rotate with an adjusted rotational amount as required.
  10. If all is well, delete the middle layer (the previous frame), merge the top and bottom layers together clipping to the image size, save the result on your clipboard for use in the next frame, save the new image back to disk, and close this frame.

All of that should take you under five minutes, two if you're on a roll. Now, multiply by thousands of frames. Good luck!

Forthcoming Resources

The first of what are hopefully several hand-assembled video sequences are now posted and available, above. Also in the works is a data set for the railroad simulator known as Trainz that re-creates the route of the Massachusetts Central roughly as it would have been when it opened for service in October, 1881; I say "roughly" because the digital elevations maps I used are from modern surveys, augmented by roadways, house and building locations from USGS topographic maps mostly dating from the 1930s and 1940s.