8.4.2 Sobel

Filter for improving the visibility of defects and removing scarcely changing background characteristics. Marks small defects which are hard to notice at the half-hue image. Marks small heterogeneities on even gray field.

Sobel transformation transforms the image each pixel of which contain an individual derivative in the respective point of the original image.

Mask of Sobel filter for marking the horizontal borders:

For marking the vertical borders a transposed matrix is used.

Filter parameters:

• Size;
• Order on Х;
• Order on Y.

The Size determines the filtration degree, that means, breadth and height of the square filter. It should be uneven and can equal to 1, 3, 5, and 7.

Order on Х, Order on Y parameters determine the value of the derivative in the filter. They usually use values 0, 1, and, 2 as maximum. Value 0 means that the derivative in that direction shall not be taken, but in this case at least one of the direction should not equal zero.

Sobel transformation combines operations of Gaussian filtration and differentiation which leads to the fact the image is more resistant to noise. Often for counting the first derivative on X or Y the following parameters are indicated:

1) Order on X: 1; Order on Y: 0; Size: 3

2) Order on X: 0; Order on Y: 1; Size: 3

Please find more information on settings and the use of filter in the step by step instruction Tutorial on the use of Sobel filter.