Edge Detection on Variance of Gray Level

Abstract

Significant subject on edge detection is that edge points with small level change can be detected without mis-detection by noise or change of gray level on shadow or curved surface.
In our study we propose a new edge detection method based on variance of gray level. This is effective method, because the variance of gray level indicates the distribution of gray level. Edge points are detected by V3 value calculated by following expression.
V3(I, J)=V2(I, J)/V1(I, J)=1/n²Σ_iΣ_j(L(I+i, J+j)-Mean)²
V2(I, J)=1/n²Σ_iΣ_j(L(I+i, J+j)-Mean)²-V1(I, J)|
For the edge points, V3 value is less than 0.65 when step change is greater than 4*σ_noise. The othen hand, for the points in the processing region without the edge line, V3 value is greater than 0.65. Therefore, V3 operator enables to detect only edge points with step change.