Introduction
This support article describes how to improve the video image quality of your cameras using Swing Catalyst's built-in video filters. The improvements may vary depending on camera type & model.
For certain camera types e.g Machine Vision, we acquire raw sensor data, this means there is no post processing done to the video images. More significant improvements can be made for these camera types. For other cameras such as web cameras or DirectShow devices image processing is done in camera and the room for improvement is not as large.
Additionally, this article also shows how the aperture of the lens makes a difference in depth of field and image sharpness.
Requirements
- Swing Catalyst version 8.0 or later.
- Any camera compatible with Swing Catalyst.
- Dedicated Graphics Card (GPU). The video filters are shader based and utilize the GPU. This can be taxing on your system (utilization can increase with 10% per camera).
Jumping straight to the results
The screenshots are from a Point Grey Chameleon 149 FPS USB 3.0 camera in our test lab, and with alternative screenshots from our offices.
Without filters:
With Sharpening, Gamma and Denoise filters enabled:
The image goes from being borderline unusable to usable.
Without filters
Sharpening, gamma and denoise filters on:
Getting to know the video filters
Now that you've seen what is possible to achieve, lets get to know the video filtering capabilities of Swing Catalyst.
The video filters can be found in the Advanced camera settings and are applied per camera.
You can access these filters by clicking the Advanced button for each camera you wish to adjust and then clicking the Filters tab.
Gamma
Machine Vision cameras usually do not apply any form of gamma correction, and by adjusting the gamma on the cameras through Swing Catalyst, there is a lot to gain. Increasing gamma will only make the dark parts of the image brighter. The black level of image changes, and too high a gamma can result in washed out (grey looking) images.
No gamma adjustment (camera default):
Gamma adjustment (1.6):
No gamma adjustment:
Gamma 1.6:
Sharpening
The sharpening filter enhances the edges of an object in the image, particularly of that what is in focus. Sharpening will make noise more pronounced, we therefor generally try to avoid using max gain when using sharpening. The amount of noise present at maximum gain varies a lot between camera sensors.
In the two screenshots from our test lab, the gamma filter is enabled as otherwise it's difficult to see the differences.
This is the default, no sharpening enabled.
Sharpening enabled:
Note how the Balance Plate and club becomes more sharp, however, the impact screen becomes more noisy.
Alternative (no sharpening):
Sharpening enabled:
Notice how the text on the power supply becomes easier to read, but conversely the image becomes significantly more noisy.
Denoise
The denoise filter is useful if there is a lot of noise in the video image, either as a result of high gain or other factors. Good results can be achieved by using the sharpening filter and denoise together.
Gamma and sharpening only:
With Gamma, sharpening and denoise:
Note how the noise around the turf and the screen has been reduced significantly.
Alternative (no filters):
Denoise enabled (notice how the chassis of the power supply has less white noise)
Sharpening and denoise enabled:
Lens correction
The lens correction filter is quite difficult to use in it's current state, and due to the fact most customers have varifocal lenses (zoom lenses) we cannot create one profile to be used. When using this filter it's important to reference straight lines in the room.
The lens distortion will always be worst at the edge of the lens, and the distortion becomes worse if you're zoomed out all the way, this is the case below.
No lens correction (note how the door frame is curved):
Lens corrected (door frame is straighter and distortion is compensated for):
It's difficult to get a fully corrected image as the corners of the image will become distorted. It's important to avoid that the object in the middle of the image (usually the user) doesn't become distorted.