AI-ordlista
Den kompletta ordlistan över AI
Region Growing Segmentation
Technique that starts from seed points and aggregates neighboring pixels based on similarity criteria (color, texture, intensity) to form homogeneous regions.
Region Merging Segmentation
Approach that begins with an initial over-segmentation (e.g., grid-based) and iteratively merges the most similar adjacent regions according to a predefined criterion.
Watershed
Segmentation algorithm that treats the image as a topographic relief, flooding basins from markers to delineate boundaries between regions.
SLIC (Simple Linear Iterative Clustering)
Over-segmentation algorithm that generates compact and quasi-regular superpixels by adapting K-Means to a 5D space (CIELAB + x,y coordinates) with a color-distance weighting.
Feature Space
Multidimensional representation where each pixel is a vector of its attributes (e.g., RGB, Lab, texture), on which clustering algorithms operate for segmentation.
Ward's Criterion
Linkage method for hierarchical clustering that minimizes the total intra-cluster variance by merging at each step the two clusters that cause the smallest increase in this variance.
Over-segmentation
Phenomenon where a clustering algorithm produces an excessive number of segments, often finer than the actual objects of interest in the image, requiring a subsequent merging step.
Kernel-based Segmentation
Clustering approach that uses kernel functions to project pixels into a higher-dimensional space where non-linearly separable clusters become linearly separable.