06 - Object Classification

Object Classification is a technique in image analysis to identify and categorize objects within an image based on their features, such as shape, color, texture, and size. In this context, an object can be defined as a set of pixels that belongs to a specific area within the image, such as a cell, tissue region, or any other structure of interest.

This approach is particularly useful in biological applications where you want to distinguish between different types of cells in a tissue, identify different cell cycle stages, classify various tissue areas, categorize different cellular structures based on their morphological characteristics, and much more.

To implement object classification, you need both the original image and its corresponding segmentation. The segmentation defines the boundaries of each object, while the features are computed from the raw image data within each segmented object area. This means that while the segmentation tells us where the objects are, the classification is based on the intensity, texture, and morphological properties extracted from the original image pixels contained within those segmented regions.

In this section, we will use the open-source software Ilastik to perform Object Classification using Machine Learning, similarly to how we used Ilastik for Pixel Classification. In this case, instead of classifying individual pixels, we will classify entire objects based on their extracted features.