Shape Detection in Computer Vision Using the Hough Transform

ISBN-13: 9783540197232 / Angielski / Miękka / 1992 / 201 str.

V. F. Leavers

Opis

Shape detection techniques are an important aspect of computer vision and are used to transform raw image data into the symbolic representations needed for object recognition and location. However, the availability and application of research data relating to shape detection has traditionally been limited by a lack of computational and mathematical skill on the part of the intended end-user. As a result progress in areas such as the automation of visual inspection techniques, where shape detection couls play a pivotal role, has been relatively slow. In this volume, Violet Leavers, an established author and researcher in the field, examines the Hough Transform, a technique which is particularly relevant to industrial applications. By making computational recipes and advice available to the non-specialist, the book aims to popularize the technique, and to provide a bridge between low level computer vision tasks and specialist applications. In addition, Shape Detection in Computer Vision Using the Hough Transform assesses practical and theoretical issues which were previously only available in scientific literature in a way which is easily accessible to the non-specialist user. Shape Detection in Computer Vision Using the Hough Transform fills an obvious gap in the existing market. It is an important textbook which will provide postgraduate students with a thorough grounding in the field, and will also be of interest to junior research staff and program designers.

Spis treści

1 Computer Vision: Shape Detection.- 1.1 Why Computer Vision?.- 1.1.1 Industrial Applications.- 1.1.2 Medical Applications.- 1.1.3 Social Applications.- 1.1.4 Military Applications.- 1.2 Why This Book?.- 1.3 Why the Hough Transform?.- 1.4 Representing Shape Symbolically.- 2 Transforms Without Tears.- 2.1 Beginning to See.- 2.2 What about Shape?.- 2.3 Tackling the Maths.- 2.4 Beginning to Compute.- 3 Preprocessing.- 3.1 The Real World.- 3.2 Spot the Difference.- 3.3 Convolution, a Necessary Tool.- 3.4 Edge Detection.- 3.5 Which Parametrisation?.- 3.6 Getting Started.- 3.7 Quantization.- 3.8 Test Images.- 4 Postprocessing.- 4.1 Results of the Transformation.- 4.2 The Butterfly Filter.- 4.3 Designer Butterflies.- 4.4 Putting Things to Work.- 4.5 Reconstruction.- 4.6 Summary.- 5 Representing Shape.- 5.1 From Lines to Circles.- 5.2 Double Houghing.- 5.3 Towards a Representation of Shape.- 5.3.1 Decomposition.- 5.3.2 Geometric and Spatial Relations.- 5.3.3 Saliency.- 5.3.4 Invariance.- 5.3.5 Stability.- 5.3.6 Accessibility.- 5.4 Summary.- 6 Which Hough?.- 6.1 Background.- 6.1.1 Historical.- 6.1.2 Whole Shape Detection.- 6.2 Refinements.- 6.2.1 Preprocessing Considerations.- 6.2.2 Postprocessing Considerations.- 6.3 Software Solutions.- 6.3.1 Computer Friendly Algorithms.- 6.3.2 Dynamically Quantised Accumulators.- 6.3.3 Constraints on Parameter Calculation.- 6.3.4 Parameter Space Decomposition.- 6.3.5 Segmenting the Image.- 6.4 Parallel Processing.- 6.4.1 SIMD Implementations.- 6.4.2 MIMD Implementations.- 6.5 Dedicated Hardware.- 6.6 The Probabilistic Houghs: A Review.- 7 A Case Study: Circles and Ellipses.- 7.1 Preprocessing the Image Data.- 7.2 The Dynamic Generalized Hough Transform.- 7.2.1 Connectivity Detection.- 7.2.2 Segmentation.- 7.2.3 Sampling of Data Points.- 7.2.4 Calculating the Parameters.- 7.2.5 Accumulation of the Parameters.- 7.2.6 Statistical Information and a Robust Stopping Criterion.- 7.2.7 Removal of Features from the Image.- 7.3 A Case Study.- 7.3.1 Edge Intensity Threshold and Connectivity Detection.- 7.3.2 Segmentation.- 7.3.3 Automatic Stopping Criterion.- 7.3.4 Results.- 7.3.5 Coping with the Unexpected.- 7.4 Discussion.- 7.5 Conclusions.- Appendix 1.- 1.1 The Radon Transform.- 1.2 Generalized Function Concentrated on a Line.- 1.3 The General Case.- 1.4 Application to an Ellipse.- Appendix 2.- Appendix 3.- Appendix 4.- References.