Research

My areas of interest include computer vision and image processing (video and image superresolution, image de-noising, compression artifacts removal), image modeling (Hierarchical Markov Random Fields), graphical models for image classification, machine learning and approximate inference (EM algorithm, loopy belief propagation) and convex optimization techniques for density extimation. Some current and past projects are listed below in order of completion (latest to oldest)

  • Linear manifold based techniques for Super-resolution.
  • Study of compression bit rates on performance of video enhancement algorithms.
  • Jpeg compression artifacts removal from images and videos[pdf].


  • Primal-dual methods for L1 regularized logistic regression[pdf].
  • Graphical models for object recognition and classification.
  • On the structure of ambiguity in reconstructing human motion over multiple frames[pdf].
    Click here to see video with ambiguous limbs Click here to see the corrected video

  • Sub-pixel motion estimation using ordinal regression in polar co-ordinates.
    Sub-pixel shift estimation

  • Image super-resolution using belief-propagation in hierarchical MRF's[pdf].
    Original License Plate Digit image restoration

  • Recognition and restoration of license plate images.