Ali El HabchiRESEARCH · ENGINEERING · TEACHING

VISION LAB / INTERACTIVE TEACHING EXPERIMENT

A different way
to see a pixel.

Choose an image, change the filter, then inspect the calculation. This demo applies a 3 × 3 kernel to a grayscale image, using the cross-correlation convention common in image-processing libraries.

Synthetic input, not drone footage. Uploaded images stay in this browser and are resized to 128 × 128 pixels.

01 / INPUT128 × 128 · grayscale

Choose a pixel with the controls below to inspect its neighbourhood.

02 / OPERATOR3 × 3 kernel

03 / OUTPUTclipped to 0–255

Signed filter responses are clipped for display. Raw values appear below.

INPUT NEIGHBOURHOOD × KERNEL

Border handling: replicate the closest edge pixel. Box blur divides the sum by 9. Pixel controls inspect interior pixels.

Small operations.
Visible consequences.

Sobel X responds to changes along the horizontal axis, making vertical boundaries stand out. Box blur averages a neighbourhood. Sharpen increases local contrast. Try the same input with each filter, then compare the coefficients and output.

More about my teaching ↗