Scale Invariant Optical Flow

Abstract

Abstract. Scale variation commonly arises in images/videos, which cannot be naturally dealt with by optical flow. Invariant feature matching, on the contrary, provides sparse matching and could fail for regions without conspicuous structures. We aim to establish dense correspondence between frames containing objects in different scales and contribute a new framework taking pixel-wise scales into consideration in optical flow estimation. We propose an effective numerical scheme, which iteratively optimizes discrete scale variables and continuous flow ones. This scheme notably expands the practicality of optical flow in natural scenes containing various types of object motion.

Aggregation of local parametric candidates with exemplar-based occlusion handling for optical flow

“…All the patch correspondences involved in AggregFlow are computed with the PatchMatch algorithm [6] based on the minimal C++ code provided by the authors 3 . A weighted median filtering with bilateral weights [65] is performed as a post-processing step to enhance motion edges as advocated in [55]. For the discrete minimization, we use available QPBO and max-flow code 4 .…”

Human motion correction and representation method from motion camera

“…The mainly approaches of motion estimation from camera involve in optical flow, frame/background difference method and object tracking. The optical flow method [8] tries to calculate the motion between two frames based on the optical flow constraint equation which assume the motion remain the same in very short time. Frame difference method needs a good and robustness background model.…”

Optical flow estimation using Approximate Nearest Neighbor Field fusion

“…Large displacement flow estimation using descriptor matching was introduced by Brox et al [4]. Xu et al [12] proposed a new frame work taking pixel-wise scales into consideration in optical flow estimation. Optical flow using rank transform was introduced by Demetz et al in [13].…”