边缘检测简介
边缘检测,简单说就是找出图像突变的地方,如上图所示。
图像微分简介
图像微分是一种简单情况的离散微分。
离散微分
连续微分是微积分的基本内容,实际离散微分概念也是类似,在微积分发明前就有了。
微分定义
d f d x = f ( x 2 ) − f ( x 1 ) ∣ x 2 − x 1 ∣ \frac{df}{dx} = \frac{f(x_2) - f(x_1)}{|x_2 - x_1|} dxdf=∣x2−x1∣f(x2)−f(x1)
d f d x = f ( x ) − f ( x − 1 ) = f ′ ( x ) Backward difference \frac{df}{dx}=f(x ) - f(x - 1) = f'(x) \quad \text{Backward difference} dxdf=f(x)−f(x−1)=f′(x)Backward difference
d f d x = f ( x ) − f ( x + 1 ) = f ′ ( x ) Forward difference \frac{df}{dx}=f(x ) - f(x + 1) = f'(x) \quad \text{Forward difference} dxdf=f(x)−f(x+1)=f′(x)Forward difference
d f d x = f ( x + 1 ) − f ( x − 1 ) 2 = f ′ ( x ) Central difference \frac{df}{dx}=\frac{f(x + 1) - f(x - 1)}{2} = f'(x) \quad \text{Central difference} dxdf=2f(x+1)−f(x−1)=f′(x)Central difference
容易看出 ( x + 1 ) − ( x − 1 ) = 2 (x + 1) - (x - 1) = 2 (x+1)−(x−1)=2, Central differnce 要除以 2 2 2
图像微分和梯度(gradient)
图像偏微分也是类似的,下面只介绍最常用的Central difference。
∂ f ∂ x = f ( x + 1 , y ) − f ( x − 1 , y ) 2 \frac{\partial f}{\partial x} = \frac{f(x + 1, y) - f(x - 1, y)}{2} ∂x∂f=2f(x+1,y)−f(x−1,y)
∂ f ∂ y = f ( x , y + 1 ) − f ( x , y − 1 ) 2 \frac{\partial f}{\partial y} = \frac{f(x, y + 1) - f(x, y - 1)}{2} ∂y∂f=2f(x,y+1)−f(