Depth completion from sparse LiDAR and high-resolution RGB data is one of the foundations for autonomous driving techniques. Current approaches often rely on CNN-based methods with several known drawbacks: flying pixel at depth discontinuities, overfitting to both a given data set as well as error metric, and many more. Thus, we propose our novel Piecewise Depth Completion (PDC), which works completely without deep learning. PDC segments the RGB image into superpixels corresponding the regions with similar depth value. Superpixels corresponding to same objects are gathered using a cost map. At the end, we receive detailed depth images with state of the art accuracy. In our evaluation, we can show both the influence of the individual proposed processing steps and the overall performance of our method on the challenging KITTI dataset.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    PDC: Piecewise Depth Completion utilizing Superpixels


    Beteiligte:


    Erscheinungsdatum :

    2021-09-19


    Format / Umfang :

    3075255 byte




    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch



    Structure-Sensitive Superpixels via Geodesic Distance

    Wang, P. / Zeng, G. / Gan, R. et al. | British Library Online Contents | 2013


    SEEDS: Superpixels Extracted Via Energy-Driven Sampling

    Bergh, M. / Boix, X. / Roig, G. et al. | British Library Online Contents | 2015


    Superpixels: An evaluation of the state-of-the-art

    Stutz, David / Hermans, Alexander / Leibe, Bastian | British Library Online Contents | 2018


    Image dehazing using adaptive bi-channel priors on superpixels

    Jiang, Yutong / Sun, Changming / Zhao, Yu et al. | British Library Online Contents | 2017


    Image dehazing using adaptive bi-channel priors on superpixels

    Jiang, Yutong / Sun, Changming / Zhao, Yu et al. | British Library Online Contents | 2017