Loading...

Image Combination for the Quality Improvement (Restoration)

Ehsandoust, Bahram | 2012

778 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 43332 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Hossein Khalaj, Babak
  7. Abstract:
  8. In this thesis, we want to improve the image quality using the auxiliary images taken by either camera networks or one camera in different times and places. This problem has been regarded so important because of the development of the camera networks and increasing the consumers' demand for enjoyment from pictures with higher quality in recent years. The problem generally consists of two different and relatively independent phases of (1) Image Registration and (2) Image Fusion. At the first step, the information in the different images are aligned and prepared to be combined using homography transforms. After that, the registered images are fused in a way depending on the target application during the second step. To this, there are two novel fusion algorithms introduced in this thesis which are both based on the inherent sparsity of the images. One of these methods uses the sparse representation of the images based on a complete dictionary (e.g. discrete cosine transform) basis, while in the other one, the Projection Onto Convex Sets scheme is employed. These methods are applied for two sample applications of Super-Resolution and Image De-Blurring and the simulation results are studied. Evaluating the results and comparing them we show that utilizing the proposed techniques leads to noticeable improvement in different quality features of the images (such as image resolution, sharpness, noiselessness, and etc.). In addition, a novel and efficient single-image super-resolution method, as one of the most significant pre-processing stages for the aim applications, is proposed using morphological filters. This technique serves a noticeable progress in regard to the traditional algorithms preserving the high frequency components and the edges of the image
  9. Keywords:
  10. Fusion Image ; Sparse Representation ; Image Registration ; Image Super-Resolution ; Image De-Blurring

 Digital Object List

 Bookmark

No TOC