End stopped based edge detector The Problem Edge

End stopped based edge detector

: The Problem • Edge detectors are not good enough. – – Sensitive to noises How to guess the parameters False positive Junctions, corners • Modern edge detectors, such as logical linear edge detector (Iversom & Zucker 1995 have better result, but there much more complicate

The Idea • The idea: stick to the biologist. • Primary visual cortex inspired, based on Hubel & Wiese work. – Detect edges by there orientation. – Use end stopped sensors. – Non-linear.

Plan End stop Edge detector • The basic Kernels 0 0 01 -1 -1 0 0 0 1 1 1 -1 -1 -1 0 0 0 0 -1 1 0 0 0 0 -1 1 0 0 0 -1 0 0 0 0 0 -1 0 0 0 0 1 -1 0 0 0 -1 0 -1 -1 0 1 1 0 0 • off & on center cell 0 0 0 1 0 0 0 -1 0 0 0 0 1 -1 0 0 0 0 0 0 0 0 1 1 0 0 -1 0 1 -1 -1 0 0 0 0 0 0 1 1 0 0 0 0 -1 0 0 0 -1 1 0 -1 0 0 0 0 -1 0 0 0

End stopped based edge detector Salt and pepper noise Edge detector based on end stopped edges no parameters. Salt and pepper noise Sobel edge detector parameter 200

Results • No parameters. • Significantly less false positive. For some of the true edges, the magnitude is profoundly higher • Less sensitive to noises. • Problem: Some of the edges are not completed or missed.

? How to improve • Larger kernel has more chances to miss edges. • Larger Kernel has less chances to “false positive”.

Improved End stopped based edge detector The pictures came from different sizes of end stopped kernels Some of the edges are missing, but appear in others

? How to improve • Still some of the edges are missing. • Combine the End stopped result with classic detectors such as Sobel.

Additive noise, Sobel edge detector Parameter 100

Results Additive noise, non-linear combine edge detector Parameter 100

Non-linear combine edge detector • • No parameters. Less sensitive to noises. More correct edges. The Edges are not “clean”.

a lot of noise • Picture with a lot of noises. • Sobel performances are poor. • End stopped edge detector is reasonable.