Previous | Next --- Slide 41 of 69
Back to Lecture Thumbnails

This is a typical 3-by-3 convolution filter. It simply takes the average on the values of the pixel values around the given pixel. The result will be blurred just like in the next slide. This is highly related to the Convolutional Neural Networks in deep learning and computer vision.


The weights matrix here is equally distributed, meaning that our convolution will resemble a box blur. This output isn't as smooth of a blur as with a Gaussian, which is sometimes more typically correlated with CNNs.