Discover published sets by community

In the Random Walks segmentation method, pixels are interpreted as nodes of a graph, and the segmentation task is formulated as a pixel labelling problem based on probabilities from random walks. Example: Segmenting an image into foreground and background by modelling it as a random walk and solving the probabilities.

Clustering techniques group similar pixels together based on their values. A popular clustering method in image segmentation is K-means, where 'K' clusters are formed by minimizing the variance within each cluster. Example: Color-based segmentation using K-means to divide an image into segments of dominant colors.

Graph Cut is a more complex segmentation approach that models an image as a graph, where pixel values represent nodes and edges represent pixel similarity. Energy minimization is used to partition the graph. Example: Extracting the foreground from the background in an image by cutting the graph into two.

This technique uses the image's histogram to find distinct peaks and valleys, which can help in isolating different regions within the image. Example: Segmenting different tissue types in a medical image by identifying peaks in the histogram that correspond to distinct intensity ranges of tissues.

This method involves dividing an image into regions based on a predicate and then merging adjacent regions that satisfy a merging criterion. Example: Splitting a landscape photograph into segments (sky, water, land) and then merging homogeneous regions to simplify the scene.

Thresholding is a simple segmentation technique that converts a grayscale image into a binary image based on a threshold value. For each pixel, if the pixel's value is greater than the threshold, it is set to the maximum value (often 255 for 8-bit images); otherwise, it is set to 0. Example: Global thresholding based on an arbitrary value, adaptive thresholding where threshold varies according to image content.

Region Growing is a procedure that starts with a set of seed points and grows regions by appending to each seed those neighboring pixels that have similar properties. Example: Segmenting an area of interest in a medical image by growing a region from a seed point inside a tumor.

Otsu's Method is an automatic thresholding technique that chooses optimal threshold value(s) to minimize intra-class variance in the image. Example: Automatically separating the text from the background in a scanned document.

Normalized Cut is a segmentation technique that partitions the image pixels into disjoint regions by optimizing global impression of dissimilarity. Instead of cutting small connections, it focuses on cutting 'weak' connections with regard to the overall connection strength. Example: Separating objects with weak boundaries from the rest of the image.

The Watershed segmentation algorithm views the grayscale image as a topographic surface and 'floods' the surface from the lowest points. Pixels with similar intensity are merged together. Example: Separating overlapping cells in a biological image.

The Level Set method is a numerical technique which evolves contours to segment regions based on certain criteria, using partial differential equations. Example: Segmenting complex structures in medical imaging by evolving the contour iteratively to adhere to the structure's boundaries.

Deep Learning-based Segmentation uses neural networks to learn and perform pixel-wise classification, enabling the identification of complex patterns in the data. Example: Autonomous vehicle vision systems using convolutional neural networks (CNNs) to segment and interpret road scenes.

Edge Detection methods find boundaries within an image by detecting discontinuities in brightness. Common algorithms include the Sobel, Canny, and Prewitt operators. Example: Detecting the edges of objects in an image to delineate shapes.

Active Contours, or Snakes, are curves defined within an image that can move under the influence of internal forces within the curve itself and external forces from the image data. They are used to localize object boundaries. Example: Outlining blood vessels in a retinal image.