Machine vision algorithms

This section describes provides a non-exhaustive list of entry points for common machine vision algorithms supported by the toolbox. Consider these as some starting points for exploration.

Linear 2D filtering

• Convolutional filtering: convolve, smooth

• Filter kernels: Gauss, DGauss, DoG, Laplace, LoG, Sobel

• Edge detection: canny

• Scale-space: scalespace, pyramid

• Gradient estimation: gradients, direction

Nonlinear 2D filtering

• Mathematical morphology: morph, open, close, dilate, erode

• Hit or Miss filtering: hitormiss, thin, triplepoint

• Rank filter: rank, medianfilter

• Distance transform: distance_transform

• Image similarity: similarity

Feature extraction

• Whole image features

  • Simple statistics: stats, hist, mean, median, std, var, etc.

  • Histogram analysis: hist, h, cf, cdf, pdf, peaks

• Region features

  • Segmentation:

    • Thresholding: otsu, triangle, threshold, threshold_adaptive, threshold_interactive

    • MSER features: MSER

    • Color k-means: kmeans_color

  • Connected component (blob) analysis: blobs

• Line features

  • Hough lines: Hough, HoughFeature

• Point features:

  • Harris corners: Harris

  • Scale-orientation invariant features: SIFT, SURF, BRISK, ORB, etc.

  • Feature matching: FeatureMatch

• Text features (OCR): ocr

• Fiducial features (AR tags, AprilTags, etc.): fiducial, Fiducial

Image retrieval

• Bag of words matching: BagOfWords

Camera models

• Central-projection (aka pinhole) camera: CentralCamera

  • Camera calibration: images2C, decomposeC

  • Pose estimation: machinevisiontoolbox.Camera.CentralCamera.estpose

  • Projection: project_point, project_line, project_conic, project_quadric

  • Epipolar geometry: E, F, epiline, CentralCamera.decomposeF, decomposeE

  • Homography: H

• Fisheye camera: FishEyeCamera

• Catadioptric (omnidirectional) camera: CatadioptricCamera

• Spherical camera: SphericalCamera

Multiview geometry

• Stereo vision: stereo_simple, stereo_BM, stereo_SGBM

• Rectification: rectify_homographies

• Bundle adjustment: BundleAdjust

Point cloud processing

• Downsampling: downsample_voxel, downsample_random

• Transform: transform

• ICP (Iterative Closest Point): ICP

Visual servoing

• Position-based: PBVS

• Image-based: IBVS