粒度计算模型及应用研究.pptVIP

A Comparison of SIFT, PCA-SIFT and SURF Weimin Liu survey INTRODUCTION RELATED WORK OVERVIEW OF THE THREE METHODS EXPERIMENTS RESULTS CONCLUSIONS FUTURE WORK INTRODUCTION Ke and Sukthankar used PCA to normalize gradient patch instead of histograms.(slow) RELATED WORK OVERVIEW OF THE THREE METHODS- SIFT detector OVERVIEW OF THE THREE METHODS- PCA-SIFT detector OVERVIEW OF THE THREE METHODS- SURF detector SIFT and SURF algorithms employ slightly different ways of detecting features EXPERIMENTS RESULTS repeatability measurement EXPERIMENTS RESULTS RANSAC, EXPERIMENTS RESULTS EXPERIMENTS RESULTS-Processing Time EXPERIMENTS RESULTS-Scale Changes EXPERIMENTS RESULTS-Image Rotation EXPERIMENTS RESULTS-Image Blur EXPERIMENTS RESULTS-Illumination Changes EXPERIMENTS RESULTS-Affine Transformations EXPERIMENTS RESULTS-Discussion CONCLUSIONS FUTURE WORK * * Key laboratory of information processing Institute of Computing Technology Chinese Academy of Science liuwm@ics.ict.ac.cn Lowe (2004) presented SIFT for extracting distinctive invariant features from images that can be invariant to image scale and rotation.(slow) Bay and Tuytelaars (2006) speeded up robust features and used integral images for image convolutions and Fast-Hessian detector.(faster) KNN – keypoint matching method (Lowe) RANSAC – find the correct match affine transformations,scale changes, rotation, blur, compression, and illumination changes (K. Mikolajczyk and C. Schmid) PCA is well-suited to represents keypoint patches but observed to be sensitive to the registration error (Some researches) four major stages scale-space extrema detection keypoint localization Orientation assignment keypoint descriptor DOG instead of Gaussian to improve the computation speed D(x, y,s ) = (G(x, y, ks ) ?G(x, y,s )) * I (x, y) = L(x, y, ks ) ? L(x, y,s ) PCA is a standard technique for dimensionality reduction PCA-SIFT uses PCA instead of histogram to normalize gradient patch The feature vector is sig