Image Processing Activity

Keywords:

Filter - A mathematical operation that is applied to an image using a kernel to alter an image’s pixel values in a specific fashion. 

Kernel - A small (smaller than the input image size) matrix of numbers used to filter an image. The kernels are applied to an image via a mathematical calculation, which is commonly a convolution. 

Convolution - The most common way that filters are applied to an image. To filter an image, the filter kernel is slid over an image and calculates the weighted sum of all the pixels under the kernel. This sum correlates to the neighborhood’s pixel value in the output image. Depending on the kernel applied to the image, the resulting output image will display a visible effect from the filtered kernel.

Background:

Understanding the variety of techniques that are at your disposal, when preparing data for processing, allows you to easily create more advanced and efficient machine learning models. Models with thoughtfully prepared data can perform a wider array of tasks and complete those tasks more efficiently.

In this activity you will be performing operations that will help strengthen your understanding of image filtering for use in machine learning applications. We will learn about how machine learning models filter images using different math operations of "kernels" and how these kernels are applied to images using convolutions, and how these convolutions and kernels work together to process an image and help you better understand which kernel algorithms to use for the specific task you are trying to accomplish. Understanding how convolution operations on kernels and input images function will allow you to choose the best combination for specific machine learning objectives.