The official DC/2 home page

The Blue Eagle (DC/2 beta) home page

Welcome to the Blue Eagle project. Nowadays all new developments have a codeword, so I'd to do the same ;-), but the official name is DC/2, or Digital Camera/2 if you prefer the long name.

First, there is no beta yet, but I've added some preliminary information about what I'm planning to do (hopefully I'll get some spare time), because I do want to read your comments on that. I do have a prototype and I've already used it, but it really is just a prototype, a the timeframe of a beta will depend on my requirements and your demand!

Introduction

Digical Camera/2 (abbrevated as DC/2) allows you to capture what is displayed on your screen in Real Time. You no longer require a video camera and digitalization hardware to create a movie to record demonstrations, material for education, show your friends the hidden secrets of your software or to simply document the handling of standard software.

You may know the phrase A picture says more than 1000 words, but unfortunately pictures are static. Now there is a solution available to create movies with software only - you just need a fast processor (>100 Mhz Pentium) and lots of temporary harddisk space (about 400 MB per minute), standard tools from MMPM/2 (to playback the video) and the OS/2 Warp BonusPack (VideoIn to create a MMPM/2 conforming video) and Digital Camera/2.

Digital Camera/2 is a technology demo at the moment, but I've created a Ultimotion video created with Digital Camera/2 in about 2 hours (to reduce download bandwidth - from 572kB to 167kB I have zipped it).

Requirements

To replay a movie created by DC/2 and VideoIn, you just need any digital video player, for example the one shipped with MMPM/2 (MPPM.EXE) or the one shipped with IBM Multimedia Viewer in the BonusPack (IB.EXE). Just ensure that your Web Client knows to play *.AVI files with the player you have installed, e.g. IBM WebExplorer will default to IB.EXE, but you can change this in seconds via the Configure->Viewers menu.

A display resolution of more than 640*480 is also suggested, as the sample video was recorded at 640*480*256. A soundcard may be useful (at least for German speakers) but is no requirement.

Technical Design

How does DC/2 work? Well, to record a video you simply start DC/2 and run it in parallel to what you want to capture. DC/2 just regularily captures the contents of the screen in Real Time, that is records exactly what you can see on your display. A single VGA display at a 256 colors resolution requires 307200 bytes video memory, this is quite an impressive data volume! Usually you can compress this to about 1/10th of the original size, but unfortunately compression takes time, and we don't have time while capturing data in Real Time.

So we have to optimize two diverging problems:

compress the captured data as much as possible, because you can't hold more than a few seconds uncomressed video in RAM and harddisk access is more than 10000 times slower then RAM access.
capture as much frames as possible in order not to lose anything that is displayed on the screen.

Therefore DC/2 had to employ an ultra fast and efficient compression algorithm, which was designed after DeltaFrame processing method. A screen content, called ImageFrame, is compared with the previous ImageFrame, all differences between them are recorded as the so-called DeltaFrame. The DeltaFrames are then saved in Real Time to the harddisk. If nothing changes between 2 ImageFrames, the DeltaFrame is empty, if something changes all changed lines are recorded in the DeltaFrame (for each differing line the different pixels beginning with the first pixel changed to the last one, aligned in 32 bit words, is saved).

But how can one get an ImageFrame? Again there are two choices:

One can use the GPI (Graphics Programming Interface) of the OS/2 PM (Presentation Maganger). The GPI is very powerful (somewhat complicated, but properly documented and easy to debug), but this has a price - the GPI is slow.
One can use - as DC/2 does - the fastest possibility, direct hardware access, and fortunately beginning with OS/2 WARP IBM has added DIVE (Direct Interface Video Extension) to OS/2. Using DIVE, DC/2 gets direct access to the physical video RAM, however as IBM could not establish XGA-2 as the graphics standard, the video RAM is partitioned in so-called memory banks.
These memory banks are windows into the whole video RAM and as their size (the aperture size) is smaller than the whole video RAM, one has to switch between the memory banks to copy the whole video RAM. Also the length of a scanline in video RAM seems to be a multiple of 2, which gives for a VGA row which has 640 pixels a scanline length of 1024, again some computational power is required (neglecting a possible bug in DIVE).

However, that's not the complete story! Modern hardware uses hardware cursors, that is the cursor you can see on the screen is not set into the physical video RAM but the graphics controller knows where the video beam generated out of the video RAM must be overlayed by the cursor. In summary, DC/2 has to emulate the cursor. DC/2 queries the cursor, gets the pixel data that the cursor is made of and sets these pixels into the pixel data read out of the video RAM via DIVE - DC/2 emulates in software what modern hardware does in hardware for performance reasons.

After recording has finished, all ImageFrames taken are decompressed out of the first ImageFrame and all DeltaFrames saved while recording. These ImageFrams are then saved as standard, compressed and numbered OS/2 bitmaps, which can then be imported to the AVI file utility of VideoIn to create an uncompressed video. The AVI file utility not only requires hundreds of megabytes temporary space, but the created video is uncompressed and therefore also takes hundreds of megabytes space per minute video.

The VideoIn recorder is then used to create a compressed video, and it really gets compressed then! It takes minutes to compress 1 minute video, but the hundreds of megabytes are then reduced to a few hundred kilobytes.

If you want to add sound to your video, just use the Digital Audio Recorder from MMPM/2 to capture your audio data, just ensure that the audio is slightly longer than the video. Again the AVI file utility is used to merge the video and audio - and ready is your new video (and you can get rid of the megabytes temporary space)!

Limitations

At the moment you can only capture when running in a 256 color video mode, because that's the only mode I know how the pixel data is organized in the physical video RAM at the moment.

Theoretically you can capture under any resolution, the higher the resolution the lower the frame rate you can get. On my PC750 (Pentium 133) I can get a frame rate of 5.5 at VGA 640*480 which gets reduced to about 1 at 1280*1024 - that's the difference between 300kB and 1.3MB video RAM!