Linux-Based 8mm Telecine
Media Conversions, my business, converts videotape and slides to DVD. My customers often ask if I also can convert 8mm film. This is the story of my adventure into converting film to DVD. There are a number of ways to make a conversion. You can run the film through a projector and use a video camera to capture the images. Although, finding a working projector is difficult. Belts and rubber drive components dry up. Worse, 30-year-old rolls of film, some with splices, may no longer stand up to the stress of being projected at 18 frames/second (f/s). Plus, most video cameras run at 30f/s and will not synchronize with the projector.
Telecines have been used since the early days of broadcast TV to convert film to video. A number of Web sites describe DIY Telecine projects (see Resources). Generally, they either rebuild a projector and use a still camera, or they utilize a flatbed scanner and a custom film transport. Based on my research, I decided to build a Telecine using a flatbed scanner. The cost of entry is low, and scanners running at 3,000dpi or above are a commodity item. You can get started on the conversion software without the film transport, and you don't need custom optics. The downside, if you're not a programmer, is that you have to write all of your own software.
I decided early in the project that I wanted to use only open-source software tools. I hosted it on an Ubuntu Linux desktop system. I knew I would need a programming language with support for scanning, serial (or parallel) port communication, a math library and an image library. A plotting and drawing library also would be helpful during program development. I also wanted a language that offered ease of programming in higher-level constructs. I was familiar with C, but did not want to use it for this project, so instead, I decided to use Python. Python is easy to learn, it's well supported by the Linux community in both on-line forums and with numerous examples of code, and error handling and type checking/conversion are part of the language. Plus, the Python Imaging Library includes an interface to SANE for scanner support.
I acquired an Epson Perfection 3490 photo scanner for the project. It has SANE drivers, a built-in backlight for film scanning and offers 3,200dpi resolution.
There are four steps to converting a roll of film: scan the film in segments, find the image frames in the segments, remove duplicate frames where the segments overlap and make a movie from the frames. I wrote three separate Python programs for the first three steps and used FFmpeg for the fourth. The software relies on cheap disk space. Frame files are copied from segment scans. Overlap removal makes a second, renumbered, copy of all of the frame files. This strategy allows each of those programs to be rerun with the same segment scans for debugging and program development.
The cost, for a 50-foot roll of film, is approximately 8GB of space for the segment scans and similar amounts of space for the log file (if debug is turned on) and each of the frame file sets. Files are written into subdirectories of the current directory and numbered sequentially. A root filename, given as a command-line argument, is used as a prefix. Scan data is written into the scans directory, and frame files are written to the frames directory. If logging is turned on, log files are written to the logs. If debug is left on (default setting), marked up copies of the scan files also are written to the logs. The markings show where the edges of the sprocket holes were found and the outline of the frame extracted. Finally, overlap-removed, renumbered frames are written to the movie directory.
The program for scanning film simply calls the SANE scanner interface, saves the scan data, advances the film and repeats for a count given as an argument on the command line. See the Film Transport sidebar for a description. You can do a project like this without a film transport, but it's tedious. Each scan takes about 80 seconds. Limits on the size of the backlight meant that I could use only about 7.7 inches out of the approximately 8.5 inches of scanner width. Allowing for overlap between the scans, a 50-foot roll of film will have about 90 scan segments and takes roughly two hours to scan.
There are two parts to this project. One involves the software that processes the scanned film and makes a movie. The other part is the design of a film transport. The film transport is the harder part of the project, because it involves creating one-of-a-kind hardware. My transport design is based on reel-to-reel tape recorders popular in the 1960s (Figure a). It feeds film from a supply reel, across the scanner and winds it up on a take-up reel. A pair of spring-loaded idlers maintains film tension. A sprocket drive advances the film.
The film transport is controlled by an embedded microprocessor. It takes commands from the Linux system over a serial port, and controls supply and take-up reel rotation and a sprocket motor for advancing the film. I was able to find both a program development and device programming environment as well as a C compiler for the Microchip PIC series of microprocessors all running under Linux. See Resources for the list of software tools used in this project.
To simplify the software, I made a film guide out of 10mm thick clear plastic film. I first aligned a steel ruler with the scanner axis, and I used GIMP to examine scans of the ruler edge. I moved it between scans until it was aligned to within approximately 50 pixels with the grid in GIMP. At 3,200dpi, 50 pixels is about 0.015 inches and more than adequate for this application. Then, I placed a piece of plastic against the ruler and glued it down with CyanoAcrylate glue. Once the glue was dry, I removed the ruler and used a piece of 8mm leader as a spacer to glue down a second guide. A sheet of glass placed over the guides keeps the film being scanned in alignment. With the film aligned with the scanner, no corrections for skewed images are necessary.
The program for finding frames actually is looking for sprocket holes. It's substituting software registration for mechanical registration of the film. Figure 1 shows a short piece of scanned film. The left-hand side is the original scan, and the right-hand side is the same scan converted to black and white (B&W).
Before we look for sprocket holes, we first find the top edge of the film. Given the alignment of the film in the guides, we could skip this step, but at this point, I'd rather not. The location of the top edge and knowing whether it's Regular8 or Super8 film (see the A Short History of 8mm Film sidebar), tells us approximately where the centerline of the sprocket holes will be.
|Containers—Not Virtual Machines—Are the Future Cloud||Jun 17, 2013|
|Lock-Free Multi-Producer Multi-Consumer Queue on Ring Buffer||Jun 12, 2013|
|Weechat, Irssi's Little Brother||Jun 11, 2013|
|One Tail Just Isn't Enough||Jun 07, 2013|
|Introduction to MapReduce with Hadoop on Linux||Jun 05, 2013|
|Android's Limits||Jun 04, 2013|
- Containers—Not Virtual Machines—Are the Future Cloud
- Lock-Free Multi-Producer Multi-Consumer Queue on Ring Buffer
- Linux Systems Administrator
- Introduction to MapReduce with Hadoop on Linux
- Senior Perl Developer
- Technical Support Rep
- Weechat, Irssi's Little Brother
- UX Designer
- One Tail Just Isn't Enough
- Android's Limits
- Free is costly
44 min 39 sec ago
- Bought photoshop CS5 for developing a website :(
1 hour 59 sec ago
- Reply to comment | Linux Journal
1 hour 48 min ago
- Reply to comment | Linux Journal
1 hour 49 min ago
- Replica Watches
4 hours 14 min ago
- Reply to comment | Linux Journal
8 hours 24 min ago
- on the path to understanding
8 hours 28 min ago
- As a fisher,we know that a
1 day 3 hours ago
- All I Say Is Worth Share!
1 day 5 hours ago
1 day 5 hours ago
Free Webinar: Hadoop
How to Build an Optimal Hadoop Cluster to Store and Maintain Unlimited Amounts of Data Using Microservers
Realizing the promise of Apache® Hadoop® requires the effective deployment of compute, memory, storage and networking to achieve optimal results. With its flexibility and multitude of options, it is easy to over or under provision the server infrastructure, resulting in poor performance and high TCO. Join us for an in depth, technical discussion with industry experts from leading Hadoop and server companies who will provide insights into the key considerations for designing and deploying an optimal Hadoop cluster.
Some of key questions to be discussed are:
- What is the “typical” Hadoop cluster and what should be installed on the different machine types?
- Why should you consider the typical workload patterns when making your hardware decisions?
- Are all microservers created equal for Hadoop deployments?
- How do I plan for expansion if I require more compute, memory, storage or networking?