From Burr to Tongue · Separating Particles from the Background

Photo and classification · red = classified as particle, blue dashed = true outlineTap to show the photo only

Detected area ÷ true area –Blobs –

Pixel-value histogram · log vertical axisred line = threshold

Threshold –Coarse-particle diameter –

Lighting

Backlight = shadows on a light pad

Light angle Elevation above the surface

Ambient light Light that fills in the shadows

Particle density Share of the frame covered by particles

Lighting gradient Brightness difference, left to right

Exposure Above 1, the background blows out to white

Threshold method

Relative to background brightness

Flatten background Divide by a blank-paper photoRemoves uneven lighting

Assumes a phone main camera at 15 cm (one pixel = 49.5 µm), blur σ 0.8 px, 2% noise, and pixel values stored as sRGB. Particle height is taken as 0.6× the diameter and coffee reflectance as 8% of the paper’s. Coarse particles are 300 µm and up. This is a model, not a real photo.