Computer Vision Foundations · free preview

The Nulling Filter

Assembling $g_t + v_x g_x + v_y g_y$ into a velocity-tuned convolution that makes chosen movers vanish.

1. Three kernels, one assembly

Three kernels, one assembly

2. Why the output is exactly zero

Why the output is exactly zero

3. A bank nulls chosen movers

A bank nulls chosen movers

4. The payoff and the open question

The payoff and the open question

5. Building the tuned kernel

You want a filter that nulls motion at exactly vx=1v_x = 1, vy=0v_y = 0 pixels per frame. Which kernel do you assemble?

6. Who disappears

A scene holds a jogger moving right at 2 pixels per frame and a dog moving left at 1 pixel per frame. You convolve with the nulling filter tuned to vx=2v_x = 2. What does the output show?

7. The velocity the filter erases

A nulling filter is assembled as h=gt+vxgx+vygyh = g_t + v_x g_x + v_y g_y and you build it with vx=1v_x = 1 and vy=0v_y = 0, so h=gt+gxh = g_t + g_x. Objects moving right at how many pixels per frame produce identically zero output?

Practice this interactively

Free account · instant grading · spaced review that schedules itself.

Start this course for free