Home » Tutorials » Image Processing » Sobel-Feldman filter: Edge Detection and Gradient Extraction

Sobel-Feldman filter: Edge Detection and Gradient Extraction

The Sobel filter (also called Sobel-Feldman operator) is an edge detection algorithm, that results in image emphasizing edges. Computer vision apps often use this image processing technique to extract the objects contours.


What is Sobel filter?

The Sobel-Feldman operator is a separable edge detection filter. It is named after its discoverers – Irwin Sobel and Gary Feldman.

Like the Prewitt operator, the Sobel filter consists of two 3×3 convolution masks. From a computational point of view, it is relatively cheap to implement.

The Sobel operator is applicable in many computer vision algorithms such as Hough transform, Harris corners detection and the popular Canny edge detector.

The Sobel masks are with such design that they provide maximum value at horizontal or vertical edge orientation. Both masks have the same coefficients as they are rotated 90 degrees to each other. Occasionally it is desirable to retrieve only one of the gradient orientation – horizontal or vertical.

Sobel filter 3x3 mask window
Sobel operator with 3×3 mask window

The image below is an example result of a Sobel operator, where we can see clear outlines of the objects.

Sobel filter
Result of Sobel operator

Preprocessing

Typically, edge recognition results are easily affected by noise in the image. Because of this it is commonly recommended to apply some low-pass filter like Gaussian.

Smoothing filters often smudge the edges as well. Therefore, many articles suggest using nonlinear edge-preserving filtering, such as the Symmetric Nearest Neighbor or Median filter.


How the Sobel filter works?

The Sobel filter works through a simple 3×3 convolution, so it is efficient for both CPU and GPU computations. The Sobel kernels are separable, which is an additional optimization option.

$\displaystyle{G}{x}={\left[\begin{matrix}+{1}&{0}&-{1}\\+{2}&{0}&-{2}\\+{1}&{0}&-{1}\end{matrix}\right]}={\left[\begin{matrix}{1}\\{2}\\{1}\end{matrix}\right]}\ast{\left[\begin{matrix}{1}&{0}&-{1}\end{matrix}\right]}$
Sobel-Feldman horizontal gradient with separable convolution
$\displaystyle{G}{y}={\left[\begin{matrix}+{1}&+{2}&+{1}\\{0}&{0}&{0}\\-{1}&-{2}&-{1}\end{matrix}\right]}={\left[\begin{matrix}{1}\\{0}\\-{1}\end{matrix}\right]}\ast{\left[\begin{matrix}{1}&{2}&{1}\end{matrix}\right]}$
Sobel-Feldman vertical gradient with separable convolution

Each image pixel is processed by each kernel in order to produce the final gradient value using equation (2). Additionally to gain performance it is acceptable to use the sum of absolute directional values as in equation (3).

gradient-magnitude-calculation
Gradient magnitude calculation

After calculating the vertical and horizontal gradients, we can also extract the edge orientation:

gradient-orientation-calc
Gradient’s orientation

We can note that usually edge detection creates thick contours of the object. In many cases, this is not very useful for object recognition and we need to apply additional processing.

To extract thin and strong edges of the image, we use the gradient magnitude and orientation. These parameters are sufficient to do edge thinning by non-maximum suppression.

Edge Detection App

You can see how the Sobel filter works in our online image analysis demo.

Gradient Non-maximum suppression

WEB APPLET

See how it works in the browser!


Source code


References