學術論文

    Image Ops

    Thursday, 21 November

    09:00 - 10:45

    Convention Hall B

    Content-Adaptive Image Downscaling

    This paper introduces a novel content-adaptive image downscaling method. The key idea is to optimize the shape and locations of the downsampling kernels to better align with local image features. In comparison to previous downscaling algorithms, our results remain crisper without suffering from ringing artifacts.


    Johannes Kopf, Microsoft Research Redmond
    Ariel Shamir, Interdisciplinary Center, Herzliya
    Pieter Peers, College Of William & Mary

    "Mind the Gap'': Tele-Registration for Structure-Driven Image Completion

    Given several non-overlapping image pieces sloppily pasted together with gaps between them, we perform a novel tele-registration method for aligning the pieces with respect to each other. Structure-driven image completion is then applied to fill the remaining gaps.


    Hui Huang, Shenzhen Institute of Advanced Technology
    Kangxue Yin, Shenzhen Institute of Advanced Technology
    Minglun Gong, Memorial University
    Dani Lischinski, The Hebrew University of Jerusalem
    Daniel Cohen-Or, Tel Aviv University
    Uri Ascher, University of British Columbia
    Baoquan Chen, Shandong University

    A No-Reference Metric for Evaluating The Quality of Motion Deblurring

    We develop a no-reference perceptual metric for automatically comparing the quality of images produced by state-of-the-art deblurring algorithms. The metric is learned based on a massive user study, incorporates features that capture common deblurring artifacts, and does not require access to the original images.


    Yiming Liu, Princeton University
    Jue Wang, Adobe Research
    Sunghyun Cho, Adobe Research
    Adam Finkelstein, Princeton University
    Szymon Rusinkiewicz, Princeton University

    Structure-Preserving Image Smoothing via Region Covariances

    We propose a novel image smoothing approach which depends on covariance matrices of image features, as known as the region covariances. Using region covariances allows to implicitly capture local structure and texture information, making our approach particularly effective for structure extraction from texture as compared to the state-of-the-art methods.


    Levent Karacan, Hacettepe University
    Erkut Erdem, Hacettepe University
    Aykut Erdem, Hacettepe University