Image processing

The above image shows the formula for the process of imaage processing. Its purpose is to improve or transform an image by processing it.

Computer vision

The above image represents computer vision. It is used for extracting the meaning and understanding the image.

Frequency domain

Visualizing the Frequency Domain of a 2D Image explains how a 2-D FFT represents image structure through spatial frequency, magnitude, phase, distance, and direction.

Edges and regions

Edge and Region Processing connects image gradients, edge linking, region features, and segmentation in one study map.

Bayes decision

Bayes Decision and Classifiers connects the MAP decision rule, the probability of error, and the Gaussian quadratic, linear, and minimum distance classifiers.

Neural networks

Traditional Neural Network MLP builds the multilayer perceptron from a single neuron, and trains it with least squares, LMS and backpropagation.

Convolutional Neural Network CNN explains how small shared filters fix the over-fitting of the MLP.

From CNN to Transformer covers attention, the pointwise feed-forward network and the Vision Transformer.

Key math ideas

Key Math Ideas explains what the main math objects mean, such as eigenvectors and the covariance matrix, and where each one is used.