Solution:

It is a method of determining the corner points or vertices of an object in the gray scale image which has only two gray levels.

Each of the interior pixels has eight adjacent pixels called its neighbors. By examining the intensity pattern of the neighbors, we can determine whether a given pixel is a corner point or vertex.

This can be done by scanning over the image with 3 x 3 corner-point templates or masks which represent all possible types of corner points.

The set of corner points templates is generated by taking the corner pattern appearing in the upper right portion of the first template and rotating it counterclockwise to generate the seven templates in multiples of $\frac{\pi}{4}$.

To search for the corner points, scan the image with the templates using the normalized crosscorrelation function. We first scan the (m- 2) (n-2) pixels in the interior of I(k,j) to find candidate pixels, pixels with a value of 1.

Once a candidate pixel is located, its eight neighbors can be calculated by cross-correlating with 3 x 3 templates.

If the normalized cross-correlation evaluates to 1; the process can be terminated since a corner-point pixel has been identified.