Overview of the Libraries

Nokia Computer Vision Library

Structure of the Library

  • The library is built on the top of Symbian OS platform, extending image capabilities of the OS.
  • Image objects provide standard OS image internals in the whole library.
  • The library provides standard image operations such as geometric transformation, filtering, color conversion, and image statistics.
  • It also provides powerful linear algebra operations for advanced image applications. The functions include creating matrices, vectors in arbitrary dimensions, and related operations.
  • There are also building blocks for future advanced libraries.
  • The library provides high-level image and video analysis functionality such as image pyramid, motion estimation.

Feature List

  • Feature categories include imaging, motion estimation, math function, and others.
  • Imaging
    • Standard operations: resize, rotate, etc; Image properties, statistics, edges, corners, features; Color conversions and manipulations; Combining of images together either with optimal quality or speed.
  • Motion estimation - determine motion of cameras (up, down, left, right, shaking).
  • Powerful math functions; Complete linear algebra library integrated.
  • Others:
    • Camera physical property measurement.
    • Standard building blocks for image processing and computer vision.
    • Functionality similar to libraries currently used on desktop.

Camera Motion Estimation Library

Melib contains an implementation of the block based dominant motion estimation technique discussed in [1][2]. Themethod is based on computation of a small set of so-called motion features which are used in consensus based robust fitting of a parametric global motion model. Software for building up this functionality is available in the release.

The original idea of the motion profile analysis used in this library was presented in [3]. First versions of the motion estimation method were developed in a project investigating image stabilization. Later, the technique was used to build user interface control for mobile devices [1]. More recent developments are presented in [2], and implementation of the global motion estimation in this release is mainly based on that paper.

  1. Hannuksela, J., Sangi, P., Heikkila, J.: Vision-based motion estimation for interaction withmobile devices. Computer Vision and Image Understanding: Special Issue on Vision for Human-Computer Interaction 108(1-2) (2007) 188-195.
  2. Sangi, P., Hannuksela, J., Heikkila, J.: Global motion estimation using block matching with uncertainty analysis. In: Proc. 15th European Signal Processing Conference (EUSIPCO 2007), Poznan, Poland (2007) 1823-1827.
  3. Sangi, P., Heikkila, J., Silvn, O.: Motion analysis using frame differences with spatial gradient measures. (2004) Proc. 17th International Conference on Pattern Recognition (ICPR 2004), Cambridge, UK, 4:733-736.