Video Resolution Enhancement

With the proliferation of smartphones, and with Ring or CCTV cameras seemingly everywhere, if there is a suspected crime committed, or an accident, or any sort of occurance of interest, there's a good chance that it was captured on video somewhere.  Unfortunately, what's captured is often compromised by low resolution, poor lighting, glare or camera shake.  These problems can all be mitigated, potentially providing valuable information.

Strictly speaking, resolution issues are not necessarily a function of the capability of the camera or the surveillance system.  Even with a 1280 x 720 pixel size, which is considered HD, an event captured at random often will occupy only a small portion of the frame.  If the area of interest is one sixth of the frame size and the video is cropped to focus on that event, then we're essentially working with video resolution of around only 420 x 360 pixels.  There just may not be enough resolution to make out certain crucial details, and poor lighting and other issues only exacerbate the problem.

Upscaling

Video resolution enhancement involves the creation of new pixels using information contained within the video.  For each available pixel, new information is gathered via the analysis of adjacent pixels and from subsequent and earlier frames.  Since new pixels have been created, the resulting frame size is inherently larger, hence the term 'upscaling.'  A 4:1 upscaling will yeild 4 times the width and height, multiplying the number of pixels by a factor of 16.

One way this can be accomplished is via AI-powered algorithms such as those employed by software designers like Topaz Labs.  These algorithms are in a constant state of development, and while the results achievable using this type of software can be remarkable, video evidence enhanced using artifical intelligence is not typically admissible in court.  To this end, Cubic Digital Intelligence has developed a patented, 'forensically valid' software system, Motion DSP, designed specifically for law enforcement and legal applications, and which we at Clarion Labs employ when appropriate.

Stabilization and Lighting Correction

Stabilization

Stabilization is exactly what it sounds like, and most any video editing software provides a feature for minimizing camera shake.  In the case of video recorded with handheld devices like smartphones, the image must be stabilized before resolution enhancement is attempted, for the reason that upscaling relies heavily on the analysis of pixels in earlier and subsequent frames.  Following the same logic, it is also important to fix a face or item of interest to a particular spot in the frame and keep it there, which normally requires manual positioning using keyframes.  The more accurately the point of interest can be kept glued to a particular spot, the better the resolution enhancement.

Lighting Correction

Conventional processing tools such as contrast, lighting and color correction, deblur and sharpening features can also go a long way towards bringing out important detail in a video.  Any number of these tools, expertly utilized individually or in tandem, can significantly improve the quality and usefulness of a video recording, potentially rendering indistinct license plate numbers or facial features indentifiable.