Motion estimation and segmentation in depth and intensity videos

Abstract

This paper investigates motion estimation and segmentation of independently moving objects in video sequences that contain depth and intensity information, such as videos captured by a Time of Flight camera. Specifically, we present a motion estimation algorithm which is based on integration of depth and intensity data. The resulting motion information is used to derive long-term point trajectories. A segmentation technique groups the trajectories according to their motion and depth similarity into spatio-temporal segments. Quantitative and qualitative analysis of synthetic and real world videos verify the proposed motion estimation and segmentation approach. The proposed framework extracts independently moving objects from videos recorded by a Time of Flight camera.

Imaging and machine learning techniques for diagnosis of Alzheimer’s disease

“…Segmentation is a major and often the first step in many image processing applications (Ghuffar et al, 2014;Gonçalves et al, 2014;Wu et al, 2014). Segmentation is commonly used to partition the image into the objects/ regions of interest and non-relevant information.…”

Physics‐based search and optimization: Inspirations from nature

“…Hosseini () applied GbSA to principle component analysis (Ghosh‐Dastidar, Adeli, & Dadmehr, ). Hosseini () applied GbSA to image segmentation of gray‐level images (Ghuffar, Brosch, Pfeifer, & Gelautz, ), where the whole image is segmented into several regions based on similarities and dissimilarities present between the pixels of the input image. Multilevel threshold is used in image segmentation, where the number of thresholds is given in advance.…”

Scene Flow Estimation: A Survey

“…Gotardo captured three-dimensional scene flow to provide delicate geometric details [31], while Liu utilized scene flow as a soft constraint for stereo matching and a prediction for next frame disparity estimation [34]. Ghuffar combined local estimation and global regularization in a TLS framework and utilized scene flow for segmentation and trajectory generation [35].…”

“…Hence, scene flow under point cloud representation [1,50,51,33,52,53,54,55,56,57,58,35,59,60] can be presented as V = (∆X, ∆Y, ∆Z) = (U, V, W ), which truly reveal the three-dimensional displacement.…”