In modern computer vision many methods are inspired by human and animal cognition. For a deeper understanding of methods used in computer vision this course connects the visual pathways of the brain and insights about our visual system from optical illusions, brain surgery, or psychophysical experiments on the one hand to the computer vision pipeline on the other hand. Biological as well as artificial vision systems can be divided into three stages: early, mid-level, and high-level vision. Among the covered topics are:

early vision:

• cognitive: brain, eye, retina, neurons; visual pathways; primary visual cortex, edge detection; spatial frequency theory, Gabor wavelets
• computational: linear filters; fourier transform; edge detection; texture analysis using pyramids, Gabor filters

mid-level vision:

• cognitive: organizing objects and scenes: perceptual grouping, figure/ground organization; object motion; self motion and optic flow
• computational: segmentation by clustering (e.g., k-means); segmentation and fitting using probabilistic methods: EM-algorithm for handling missing data problems; tracking with linear dynamic models, Kalman filtering

high-level vision:

• cognitive:constancy and illusion; theories of object categorization
• computational:finding templates using classifiers: feature selection with principal component analysis; recognition by relations between templates

The course is concluded with the problem of finding images in digital libraries as an example for recent computer vision applications.

As this course is intended for students enrolled for the Master's degree on ''Robotics and Automation'' it is held in English. Of course, other students are also very welcome to attend the course! (Only the computational part of early vision will partly overlap with the course on ''Digitale Bildverarbeitung'' in the last summer semester.)