A detailed, story-driven companion to Richard Szeliski, Computer Vision: Algorithms and Applications, 2nd edition. Mira, a master’s student, builds a delivery robot’s perception system from first principles. Every chapter begins with a concrete failure, develops the geometry or algorithm, works a case, defines evaluation, exposes common mistakes, and ends with an exam-ready study sheet.
Why vision is inverse inference, how geometry and recognition interact, the field’s major tasks, and how a scientific vision system is evaluated.
chapter 2 · FoundationsProjective geometry, lenses, calibration, radiometry, color, sensors, noise, and the assumptions hiding inside a digital image.
chapter 3 · FoundationsPoint operations, filtering, Fourier analysis, pyramids, warping, color transforms, denoising, and restoration.
chapter 4 · Core MethodsLeast squares, robust losses, RANSAC, probabilistic inference, gradient methods, sparse structure, and uncertainty.
chapter 5 · LearningNeural building blocks, convolution, backpropagation, optimization, architectures, training practice, and failure analysis.
chapter 6 · LearningClassification, detection, segmentation, pose, instance recognition, metrics, data bias, and modern recognition pipelines.
chapter 7 · CorrespondenceEdges, corners, invariant regions, descriptors, nearest-neighbor matching, geometric verification, and learned features.
chapter 8 · CorrespondenceParametric motion, homographies, direct alignment, bundle refinement, panorama geometry, blending, and ghosting.
chapter 9 · VideoOptical flow, tracking, layered motion, regularization, coarse-to-fine estimation, evaluation, and motion boundaries.
chapter 10 · ImagingHDR, denoising, deblurring, super-resolution, matting, compositing, active illumination, and mobile imaging.
chapter 11 · GeometryEpipolar geometry, camera pose, triangulation, bundle adjustment, loop closure, maps, and drift.
chapter 12 · GeometryStereo correspondence, cost volumes, regularization, multi-view stereo, active depth, monocular cues, and metrics.
chapter 13 · GeometryVolumetric fusion, signed distance fields, surfaces, meshes, texture, neural fields, and reconstruction quality.
chapter 14 · SynthesisView interpolation, light fields, layered representations, environment matting, novel views, and neural rendering.
chapter 15 · SynthesisHow geometry, learning, physics, data, optimization, and applications fit together—and what remains unsolved.