Analyzing Video With iNPUT-ACE

In early June, Joshua and I attended a digital symposium put on by the creators of iNPUT-ACE. The ten sessions covered many of the nitty gritty details about the video analysis process, and highlighted some ways that iNPUT-ACE's software is tackling these issues. It's great to have another tool that helps us be more effective and efficient in our drive to turn raw evidence into compelling visual stories.

What is iNPUT-ACE?

iNPUT-ACE is a video investigation software tool that is becoming a one stop shop for all things video analysis and compositions.

How does it help Fat Pencil Studio?

1. Efficiency. The biggest thing that iNPUT-ACE helps us with is being more efficient with our time. Most surveillance video is stored in a proprietary format that only a very specific software can open. There are over 1,000 proprietary formats and their corresponding software does not always work well. There can be software compatibility, display, or playback problems (ie. incorrect aspect ratio / frame dropping). It takes time to process through these issues to get to something usable that will be helpful for the client. iNPUT-ACE makes bypassing these proprietary format possible.
2. Metadata. iNPUT-ACE does a great job preserving information about the raw video file. Knowing the proper aspect ratio of a video can help determine is a video is being squashed or stretched, and knowing exactly what the frame rate should be helps be more certain about timing. When we are able to get this information it's easier to understand how what we see in the video matches with reality. Analyzing the metadata is also helpful in cases where we suspect the video footage is not original, because DVR playback has been recorded with another camera.

Interesting things discussed at the seminar

• Infrared cameras affect color and race perception. Often when video is taken at night, it is infrared. This makes it possible to make out details in darkness. The downside to this is that it is not able to detect much color information. So it can be unreliable when trying to determine the color of clothing or other objects, this deficiency also can effect the appearance of race. Below is an example of the stark difference in the view of the same object under normal lighting conditions vs seen with infrared light.

• Rolling shutters can distort time. Traditional cameras have a global shutter which exposes every element of the image sensor at the same time. However, some video cameras have a rolling shutter which progressively exposes the image sensor from top to bottom, creating a situation where the things at the top of a frame are happening slightly before what is seen at the bottom of a frame. This is not usually a big deal, but can cause strange artifacts with rapidly moving objects. For example a single video frame might show two muzzle flashes that did not actually occur at the same time.

• Interpolation. When attempting to zoom & enhance video, additional pixels are created in order to display a larger image. This process requires interpolation, and there are different algorithms that can be used. Nearest neighbor, bilinear, and bicubic are three of the most common.
  • nearest neighbor is ugliest, but best for representing actual pixels, so long as you use even increments of scaling- 200, 300, 400%
  • bicubic is good for maintaining smooth edges
  • bilinear is good for maintaining edge details in applications where measuring is important

• Lens distortion. The wide angle lenses typically used in security cameras create significant distortion, especially at the edges of the image. This means edges that are straight in real life may appear curved in the video footage. We often use photoshop to deal with this, but iNPUT-ACE is also able to assist in correcting lens distortion. This is becomes very important when using camera matching or photogrammetry to measure the position of objects that appear in the footage.