- The first task was about the converting the Cartesian coordinate into homogenous coordinate, and finding the point of intersection of coordinates in homogeneous coordinate.
- Image de-noising: The second taask is about image de-noising using Gaussian, Median and Bilateral Filtering. This was done by calculating the singular value decomposition (SVD) and then calculating the 1D convolution for 1D horizontal and vertical filters. The resulting 1D Filters are then able to reconstruct the original 2D Filter by simply matrix multiplying them together.
- Hybrid images: In this task a hybrid image that combines facial images of a wolf and a man was constructed. In addition, log magnitudes of the Fourier transforms of the original images and their low-pass and high-pass Filtered versions was visualised.
- The second week of the computer vision exercise covered the implementation of a small neural network with one hidden unit as first task and
- Observation of different building blocks of a convolutional neural network as a second task.
*** The tasks are implemented in MATLAB for this week.