Spark Convolution Patch Save

Blur, sharpen, edge-detection, and convolution patches for Spark AR

https://en.wikipedia.org/wiki/Kernel_(image_processing)

All of the patches have a strength input, which can be controlled in the demo by clicking on the controller block and using the strength slider in the properties panel.

TIP! If you are using gaussian blur, a more performant option is to chain two directional blurs together (one horizontal, one vertical).

Patches

Numbers in the patch names signify the size of the kernel that is used. Lower is better for performance, higher is better for quality.

BlurDirectional3, BlurDirectional5

Blur that accepts a vector for directional blurring. Direction vector is normalized, so any range of numbers is acceptable

Convolve3, Convolve5

General purpose convolution patches that are used as a base for the other patches.

UnsharpMask5

Really good looking sharpening. Just wow. Great job.

Sharpen3

Harsh sharpening, good for enhancing small details.

BlurGaussian3, BlurGaussian5

Gaussian blur. You know the one.

EdgeBox3

Boxy edge detector.

EdgeCross3

Crossy edge detector.

Edge detection tutorial (outdated)

Resources

Learn more stuff by watching my Spark AR Tutorials on YouTube!

Follow me on Instagram @positlabs and try out my effects!

Browse my open-source Spark AR repositories on Github!

Have questions? Join the Spark AR Community group on Facebook.

Donations

If you used this in client projects, or simply enjoyed making effects with my open-source projects, please consider a donation or sponsorship. One-time donations can be made with PayPal. Subscriptions can be through PayPal or Github Sponsors (click the heart sponsor button at the top of the page).