Testing Ikena's Video Enhancement SoftwareBy: IPVM Team, Published on Mar 16, 2010
Driven by Hollywood sensationalism, public interest and expectations for miraclous video surveillance enhancement is common -- unrealistic and dangerous. A popular YouTube video captures the essence of Hollywood's approach:
By contrast, we are interested in what 'real-world' video enhancement software could perform.
To that end, we tested MotionDSP's Ikena video enhancement software [link no longer available].
Key findings include:
- With modest on-line training, fairly simple to use but optimization is more 'art' than science
- Limitations on what scenes/scenarios allow for meaningful security enhancements
- Strong enhancement for high noise / low light scenes
Ikena is a video enhancement program by MotionDSP. Ikena is designed to work with suveillance video as well as consumer video sources and UAVs.
Ikena's algorithm eliminates noise by comparing adjacent frames in a video clip and performs resolution, focus, color, and brightness/contrast enhancements to extract details of a specific subject. By contrast, image programs like Photoshop can only interpolate image data from a single frame.
The Ikena software does not integrate with VMS/DVRs: Exported VMS clips must be manually imported into the software, then exported again by the investigator. In addition, Ikena should be able to play video files that run in Windows Media Player. Ikena does not support proprietary video codecs from VMS manufacturers.
Additional features include the following:
- Installs on Windows based PCs only.
- Wide codec support for video playback.
- Uses significant video processing resources, dual core processor is required, Quad core processor recommended.
- Ikena has a version that supports NVidia based video cards which will offload the processing to the dedicated GPU.
- Software is distributed via authorized resellers.
Best Fits for Using Ikena
When using Ikena with surveillance video, the following best practices are critical to consider:
- Since Ikena depends on a series of frames where a subject is moving minimally, fast moving objects in surveillance video are likely to cause problems. Examples include a person running or a car driving quickly through a scene. If at any point, the target stops briefly, that will provide the best opportunity for enhancement.
- Surveillance video is captured over a wide range of Field of Views (FoV). When a FoV is too wide, Ikena is unlikely to be able to provide meaningful benefits for security use. At very wide FoV (over 100 feet), it is unlikely that Ikena will provide any improvements simply because the number of pixels per object is too low.
- Surveillance video is captured with a wide range of resolutions (from 320 x 240 to 1920 x 1080 are realistic in practice). When the resolution is too low, Ikena is unlikely to be able to provide meaningful benefits for security use.
- Pixels on target (which is a combination of FoV and resolution) are a key metric in projecting whether Ikena can provide realistic benefits. For example, at an 18 foot wide FoV with CIF resolution, no material benefits were demonstrated (face was blurry in both). However, at a 9 foot wide FoV with CIF resolution, material benefits occurred (more details of the face were displayed).
- With standard resolution cameras, we believe it is unlikely that any material benefits will be possible in scenes greater than 50 foot wide FoV. This is a rough assessment. However, users should set appropriate expectations when dealing with wide FoVs or lower resolution images.
If you want background and details on using Ikena, we recommend viewing their on-line training videos [link no longer available].
The following video highlights how a security investigator can import video clips into the software and the techniques used to enhance the video. The following highlights should be noted:
- Default presets are used as a starting point, not a complete solution for enhancement.
- "Matching Frames" sets the amount of frames analyzed before and after your point in the timeline.
- The "Fusion Level" slider is adjusted depending on how much the subject is moving in a scene.
- Adjusting deblurring settings too high will degrade your image. For the "Focus" slider, Ikena recommends adjusting the range from 0.25 - 0.75.
- Moving your region of interest and point in timeline will further adjust image quality.
- Merged video output is limited to VGA clips.
Image Quality / Enhancement Analysis
We conducted a competitive comparison between:
- Original 4CIF still image from a surveillance
- Manual enhancement using GIMP image editor of that surveillance still
- Ikena enhanced image of that surveillance still
- An original 1.3MP still image with the same FoV as the 4CIF still that was enhanced
We repeated the same type of test with a high noise / low light scenario and achieved the following result:
We discuss this and other quality concerns in the screencast below noting:
- No material difference in enhancements with an 80' wide FoV image.
- Moving subjects are difficult to enhance.
- Switching to video clips with a 9' wide FoV yielded more material details on the subject's face.
- Ikena enhanced a 0.1 lux low light scene at a 9' FoV and it filtered out significant camera noise and facial features can be identified.
In the best case scenarios, Ikena can provide modest improvement in image quality that can reveal material details to an investigation. However, it will not match deploying new higher resolution cameras nor will it come anywhere close to Hollywood 'infinite zooms'. In all deployments, integrators and users should appreciate that fast moving objects and low pixel on targets scenes are unlikely to provide material security benefits.
Given its price point of $7,000 MSRP per seat, it makes sense for users with regular investigation/cases (usually a few dozen cameras or more or an organization with distributed sites). In the scenarios, rather than upgrading cameras (whether from 4CIF to 720p or 720p to 1080p, etc.), it can be significantly less expensive to use the software and provide some but not all of the benefits of higher resolution. In these scenarios, when needed, video can be exported from the VMS system, sent to an investigator and the investigator can enhance the video/image.