Loading...

Feature-Based Registration of Brain MR Images

Abbasi Asl, Reza | 2013

561 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 44076 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Fatemizadeh, Emadoddin
  7. Abstract:
  8. Efficient feature selection from images would lead to a lower operational cost in image processing procedures. Recently, alpha stable filters have been used to extract significant features from medical images.
    A Feature-based method is introduced for image registration. The feature extraction is based on Bidimensional Empirical Mode Decomposition (BEMD). We have proposed two approach. The first one is a feature selection and reduction method. The registration is performed based on selected features and Bspline interpolation. The second approach is a hierarchically method based on BEMD features. Both methods improved the registration based on intensity of the images, especially for the images with Bias Field. The improvement is in convergence ratio is 25% for the first approach and 17% for the second one.
    In this Thesis, a new approach is introduced to select and reduce the dimension of extracted feature vectors of images. This approach which we have named it Maximum Minimum Response Orientation (MMRO) is based on the selection of the maximum and a minimum response of the alpha stable filters over the features with different orientation in frequency domain. This algorithm selects the rotation invariant features which are suitable for image registration purposes. Two discriminating value are introduced and utilized to evaluate the efficiency of the algorithm. First approach is based on the HAMMER similarity measure (HDV) and second one is based on the direction of feature vectors (DDV). Discriminating ability of the features selected by the proposed method is compared with Maximum Response orientation (MRO) method and an average improvement of 15% is obtained in HDV value and 13% in DDV value
  9. Keywords:
  10. Feature Extraction ; Image Registration ; Bidimentional Empirical Modes Decomposition ; Alpha Sable Filter Bank

 Digital Object List

 Bookmark

No TOC