|
Articles |
 |
|
|
|
Articles |
|
| Were the Pentagon videos manipulated?
Mark Sugrue and Russell Pickering Preface I received the original .mpeg files directly from the Department of Justice via www.flight77.info/. They were 32 megs each as opposed to the approximately 5 megs of the .wmv file from the DoJ site. I sent them to my friend, Mark Sugrue, who did the original video analysis from the Google .avi. Here is the updated review and 4 image frames. The two of the "plane" were originally 352x240 .bmp files at around 250 kb. I put them in Photoshop CS and increased them by 200% and saved them as maximum .jpeg's. The police car is from a smaller original. Russell Pickering
Abstract
Simple analysis of the structure of the recently released pentagon video shows that it has been extensively processed and manipulated. Evidence suggests that the original source video was produced by an analog CCTV camera, likely running on NTSC standard at 30Hz and 525 interlaced lines of pixels. This was being recorded at about 1-2 frames per second onto an analog format. After the event, evidence suggests that the video was digitized into 180 frames, reduced in resolution by at least 50%, deinterlaced and then selectively recombined. This digitally manipulated output video was then re-recorded back onto an analog media (such as VHS) which resulted in the addition of multiples of each frame and extra video noise. This final analog version was then re-digitized and released to Judicial Watch, resulting in a low quality, noisy video. INTRODUCTION
In this discussion I will only be talking about the totally new video, rather than the older one of which 5 frames were already in the public domain. Flight77.info received the videos on CD-ROM in MPEG format. I am not going to talk about the content of the video at all, just its structure and how it was processed. FEATURES OF THE VIDEO
The video is constructed of 5 757 frames of pixel size 352 by 240. Oddly, every 32 frames are nearly identical. I say nearly, because while each group of 32 frames certainly contain no movement and were certainly derived from the same original CCTV output frame, by running the video through frame differencing you can easily detect small jpeg artifacts and noise within the 32 frame groups. Further, the video occasionally shifts up or down by a tiny amount, in a seeming random fashion. Finally, there is a faded black border around the frames. These are all clues to the probable history of this video, and how it was constructed. PROBABLE ORIGIN
Most CCTV cameras in the US produce a very similar output to regular TV video format, and despite advances in digital technology, most are still analog. The camera position here seems to be in a non-critical role at the entrance to a staff car park. This, coupled with the faded black boundary indicated that this video was probably produced by a regular NTSC interlaced signal recorded onto an Analog medium such as a VHS cassette. This analog video would then have been digitized, processed, and released to Flight77.info on the CD-ROM. NTSC is interlaced, meaning that alternate rows of pixels are refreshed at twice the quoted frame rate. So that, with a NTSC frame rate of just under 30 fps, odd and even pixels rows are alternately refreshed at just under 60Hz. To save videotape, Analog cameras only record one frame every second or so (i.e. every 30th frame), and that seems to be the case here. (Modern digital CCTV systems commonly record at a variable frame rate, depending on activity in the scene. The fact that this did not happen here is further evidence that the system is an older Analog CCTV) When digitizing the video, you have the choice of deinterleaving the
frames to produce a video of 60Hz, but with half the vertical resolution.
One disadvantage of this is that because alternate frames are produced
by pixel positions with a slight vertical offset, the digitized video will
vibrate up and down in an annoying way. However I believe that this video
was produced in this way because at irregular intervals, the video can
be seen to shift up and down by approximately one pixel. The irregularity
indicates that the video's producers cherry-picked certain frames from
either deinterlace output, and built them into one video. Figure 1 shows
how the original video may have looked before it was deinterlaced.
Fig 1. The appearance of interlaced frames depends on the amount of motion in view. In normal circumstances, there is little visible difference between an interlaced frame and a progressive scan frame, with only some tell-tale 'hairs' surrounding moving objects, as seen in the simulated interlacing on the left hand image. In the right hand case, where the explosion causes a massive and sudden appearance change, the interlaced effect is quite profound. DIGITIZED, BACK TO ANALOG, AND DIGITIZED AGAIN?
Finally, the Department of Justice video is constructed of 180 original frames, each reproduced 32 times. The CCTV camera would have recorded frames to analog at a rate of 1Hz, but each frame only once. Why would the Judicial Watch video have frame multiples? Why would there be intraframe noise within these 32 frame groups? Digital video formats can have any desired frame rate. There is no need to have multiple frames, when the computer can simply display a particular frame for an arbitrary length of time. Analog cannot do this however. On VHS, if you want a frame to display on screen for 1 second it must be reproduced on tape a number of times. I propose that the original 180 CCTV output frames were digitized, deinterleaved, possibly manipulated, recombined into one video and then recorded back onto a VHS (or other analog format) cassette. This version was then digitized again and released to Flight77.info. Two further points of evidence of this unusual processing history is 1) that the video contains 179 frames repeated 32 times plus one frame repeated only 29 times - indicating that the digitization process was likely switched off by hand. And 2) this camera is at the entrance to a car park. It is aimed at the incoming vehicles and is probably intended to record their license plate numbers. However, the police car license plate is not visible. In fact, as similarly fine detail is visible else where in the video, I suggest it has been digitally blurred. This blurring is identical in all the 32 multiples of this frame. This must have been done after the video was first digitized into 180 frames, but before it was copied back to VHS. CONCLUSION
The result of this, whether intentional or not, is that the resultant video is of poorer quality, lower resolution, and contains more noise than the original. The important point is that this video has a long history. I suggest that the Pentagon be asked to release the original, analog and unedited version of this video, along with the 84 other videos they hold. ABOUT THE AUTHOR
Mark Sugrue a researcher in Royal Holloway, University of London, soon to complete a Ph.D. in the Machine Vision of CCTV cameras. FURTHER READING _______________ Wikipedia: Interlacing, NTSC, CCTV, AVI, JPEG
|
|
