Robust and computationally efficient online image stabilisation framework based on adaptive dual motion vector integration. Issue 5 (28th June 2019)

• Record Type: Journal Article
• Authors: Yousaf, Adeel; Hanif, Muhammad Shehzad; Khan, Muhammad Jaleed; Iqbal, Mahboob; Khurshid, Khurram
• Is Part Of: IET computer vision. Volume 13:Issue 5(2019)
• Page Start: 461
• Page End: 468
• Publication Date: 2019-06-28
• Subjects: cameras -- video signal processing -- feature extraction -- smoothing methods -- image motion analysis -- vectors; adaptive dual motion vector integration -- nonstatic platforms -- intentional motion -- two-dimensional image stabilisation technique -- novel adaptive motion smoothing technique -- motion vectors -- vehicle-mounted cameras -- online image stabilisation framework; Computer vision -- Periodicals; Pattern recognition systems -- Periodicals; 006.37
• DOI: 10.1049/iet-cvi.2018.5368
• Languages: English
• ISSNs: 1751-9632

Abstract

Abstract : Image stabilisation aims to compensate and smoothen the effects of undesired trembling motion of cameras mounted on non‐static platforms. It becomes quite a challenging task in the case of moving platforms, such as ground vehicles, unmanned aerial vehicles, and handheld devices. Many satisfactory solutions to the image stabilisation problem are proposed in the recent literature, but most of these methods are not adaptable for handling a wide range of intentional motions with minimum lag, especially in real‐time scenarios. In this study, the authors propose an online two‐dimensional image stabilisation technique based on dual motion vector integration, which is a novel adaptive motion smoothing technique that employs an average length of motion vectors to estimate the intentional motion. The overall computational cost of the proposed system is significantly reduced by employing frame‐shaking judgment that only allows processing of jittering frames. Promising experimental results have been obtained on challenging videos obtained from hand‐held and vehicle‐mounted cameras which demonstrate the robustness and effectiveness of the proposed technique against feature point mismatching and presence of moving objects within the scene at a frame rate of 30 frames per second.