2018/05/22 by Janne Mustaniemi, Mustaniemi, Janne, Juho Kannala +7
Computer Science · Engineering · #Advanced Image Processing Techniques #Advanced Vision and Imaging #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Image Processing Techniques and Applications
paper · pdf · doi:10.48550/arxiv.1805.08542
openalex publication_date 2018/05/22 · openalex created_date 2022/10/02 · openalex updated_date 2026/07/28
Many computer vision and image processing applications rely on local\nfeatures. It is well-known that motion blur decreases the performance of\ntraditional feature detectors and descriptors. We propose an inertial-based\ndeblurring method for improving the robustness of existing feature detectors\nand descriptors against the motion blur. Unlike most deblurring algorithms, the\nmethod can handle spatially-variant blur and rolling shutter distortion.\nFurthermore, it is capable of running in real-time contrary to state-of-the-art\nalgorithms. The limitations of inertial-based blur estimation are taken into\naccount by validating the blur estimates using image data. The evaluation shows\nthat when the method is used with traditional feature detector and descriptor,\nit increases the number of detected keypoints, provides higher repeatability\nand improves the localization accuracy. We also demonstrate that such features\nwill lead to more accurate and complete reconstructions when used in the\napplication of 3D visual reconstruction.\n