|
|
| Laura Teodosio, President and CEO, Salient Stills |
| by GV staff |
|
July 7, 2008
|
|
|
Q: With so many image clarification tools out there, what seems to be the most popular feature?
A: We have seen a clear trend toward the request for features that work with digital video, particularly digital video as captured from DVR systems. While digital video still has many of the same problems that exist in analog -- such as poorly exposed and resolved video due to poor lighting and camera placement and maintenance -- there are many new problems seen in digital video, including compression artifacts and proprietary video formats.
Our most popular feature is screen capture that captures a proprietary video and converts it to a non-proprietary form for sharing and analysis. There are many freeware screen capture tools available, but not all take into consideration the issues that are important to law enforcement, such as properly capturing the overlay information which may not correspond so cleanly with the video images -- and properly handling the timing of these movies once the capture has been done. You need to ensure that the resulting movie accurately represents when the image samples were captured.
For those still working with analog video, automatic demultiplexing is still a very popular and often requested feature. Without a smart demultiplexing algorithm, it could take days to split out cameras from an analog multiplexed tape.
***image1***
Q: How important are standard operating procedures for law enforcement?
A: Important. Like any other evidence, there are many points in the process where the video and its image contents can be compromised. All participants in the legal process need to feel that any video evidence has been handled and processed properly.
Law enforcement agencies have a long history of understanding the need for and creating standard operating procedures, and are applying that expertise to the field of FVA. For those in need of pointers, there are dedicated groups of law enforcement professionals working to create and share the best practices for video SOPs, in particular, SWIGIT/SWIGDE.
The SOPs should be written with the intention that the result can be reproduced by someone else. They should also, with a sound scientific basis, characterize the methods to be used. They should not define a forensics video analysis operation so tightly by requiring a particular vendor's tools and settings for the operation. That only stifles innovation in the field.
Q: How do tools like VideoFOCUS turn video of an unreadable license plate into a reasonably clear image?
A: In some instances FVA can do this. In most instances it can't, particularly in situations where the plate has been recorded only over a small area of the image. But there are two techniques -- frame averaging and super-resolution -- that can be applied in instances where a plate has been imaged over a fair number of pixels. The idea in both cases is to use the fact that a particular image and the objects in it have been recorded over and over again on video, typically 60 times a second in analog video.
Frame averaging is a technique of noise reduction
. Video noise, or snow, is introduced by the electronic components in a video capture environment. Video noise is not stationary; it moves to a different location in each frame. By looking at a particular pixel location over time, we can ascertain the most faithful recording of that area.
Super-resolution is a technique to increase resolution as well as reduce noise. It uses the motion inherent in the image pairs to better align the images. It merges several of these low resolution images are used to create a higher resolution image.
Q: Shows like CSI distort the public's perception of forensic video analysis. What are the limits of the technology?
A: I don't watch CSI anymore or at least the parts where they do FVA. Unlike on CSI, FVA cannot get a clear read on a number plate from the reflection in someone's eyeball ... at least not today.
Basically, we can't create image data that was not originally recorded. On top of that, some key details in the images are also thrown away in any compression process that the video may undergo -- like what happens in a DVR system. By the time the video arrives on an investigator's desk, there may not be much to work with. But what there is can be adjusted by image processes that attempt to make the recorded information more legible to the human brain.
Q: What's a better original source material for forensic video analysis, analog videotape or digital video files? Why?
A: This is not an either-or question. Each has its strengths and weaknesses. Digital has the promise to be a much better recording environment, but many DVR systems are set up in a way that creates an inferior recording to what analog would have provided with the same camera. This is mostly due to settings on the DVR -- high compression ratios result in small frame size and low frame rate. The data loss introduced by these settings can be significant and is not recoverable.
Often, a full frame rate analog system with well maintained cameras and decks will outshine a highly compressing digital system. The challenge is to educate the users of DVR systems so that systems are set up to record as much of the original video signal as possible.
Q: How prominent are cell phones and other mobile media devices becoming in forensic video, and how effective can FVA tools be with these sources?
A: These mobile devices are becoming quite prominent. We are seeing an increased demand for the ability to work with these kinds of videos. The first challenge is to get the video off the particular device. There are a number of software tools available to do this. Then, the FVA processing can begin.
Cell phone video is typically low data rate, meaning low resolution per frame and low frame rate. And it still suffers from many of the same problems as other video. It may be improperly exposed, it may be encoded in proprietary formats, it may need to be edited, cropped, zoomed, etc. In the process of capturing and converting cell phone data from the phone, be sure that the tool properly and faithfully retained the timing of the information of each individual video frame.
| COMMENTS (1) | | 01/21/2010 | | I would like to take this opportunity in providing some thoughts concerning Laura Teodosio's responses. First off, there are only a handful of PA's or developers world wide whom have created the ground work for FVA soft based solutions in the last decade or so. Laura Teodosio is certainly one very prominent contributor, and in some specific cases has raised the bar for results oriented turnkey systems. Unfortunately, at present forensic image analysis has reached a road block when considering aggressive discovery results. There are only but a few forensic clarification products in the industry which ultimately provide both similar and unique benefits.
While maintaining a direction of innovation and development, there needs to be additional sciences implemented to compliment forensic identification. For example, a significant portion of identification involves lic. plate identification. In this respect given the limitless variances concerning image acquisition, moving forward a supportive technology for example would be the implementation of RFID's in the actual plate itself. Much like having a red light camera at an intersection, there would also be an RFID receiver which would record the actual digits on the plate. This may seem as an intrusive tool, however given the current methods used it would be no different in purpose aside from a more accurate identification resource. To further this solution, based on example - the image may also be resolved at the point of acquisition while utilizing other technologies in tandem.
Any comments regarding this post can be forwarded to:
nicholas.guanti@yahoo.com |
|
|
|
More...
|
|
|
|
|
|
|
|
|
|