eduzhai > Applied Sciences > Engineering >

Non-parametric spatially constrained local prior for scene parsing on real-world data

  • Save

... pages left unread,continue reading

Document pages: 22 pages

Abstract: Scene parsing aims to recognize the object category of every pixel in sceneimages, and it plays a central role in image content understanding and computervision applications. However, accurate scene parsing from unconstrainedreal-world data is still a challenging task. In this paper, we present thenon-parametric Spatially Constrained Local Prior (SCLP) for scene parsing onrealistic data. For a given query image, the non-parametric SCLP is learnt byfirst retrieving a subset of most similar training images to the query imageand then collecting prior information about object co-occurrence statisticsbetween spatial image blocks and between adjacent superpixels from theretrieved subset. The SCLP is powerful in capturing both long- and short-rangecontext about inter-object correlations in the query image and can beeffectively integrated with traditional visual features to refine theclassification results. Our experiments on the SIFT Flow and PASCAL-Contextbenchmark datasets show that the non-parametric SCLP used in conjunction withsuperpixel-level visual features achieves one of the top performance comparedwith state-of-the-art approaches.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×