Loading...
Search
Search in this resource
sort by
Send to Friend
 Digital Object List
 Bookmark
 Review List
  1. انطباق تصاویر MR تحت تغییرات شدت روشنایی

MR Image Registration under Variant Illumination

Shirpour, Mohsen | 2014

2318 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 46082 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Manzouri, Mohammad Taghi
  7. Abstract:
  8. Image registration is defined as matching two or more than two images which are taken from a same scene in different times, from various views and using different sensors. By matching, we mean finding the transformation function between the images. Image registration
    has many application in medical domains such as analyzing changes in body limbs, supervision of the tumors growth, and robotic surgery. The challenges of image registration are unavailability of the transformation function, existence of the outliers in the images, changes in the images intensity, changes in the image geometry, noise, etc. Image registration consists of four steps: (i) feature extraction, (ii) choosing a similarity measure for finding the corresponding points, (iii) estimating transformation function, and (iiii) applying the estimated transformation function on the images. Based on the feature space, image registration can be classified into three classes region based registration, feature based registration, and combined feature-region registration. Illumination changes which are very common in MR images has resulted in low accuracy of the region based registration class. In this research, we aim at improving the region based registration algorithms which are based on images differences. Hence, we have introduced new region based registration algorithms that can deal with illumination changes in the images. In the first proposed method, we have used a filter for removing the noise form the DCT coefficients. The second proposed method assigns weights to the images residual pixels. More precisely, it assigns low weights to the regions with high residual and vice versa. We have tested the proposed methods on some medical images. The results have demonstrated that the proposed methods are more robust and accurate than the state of the art region based registration methods
  9. Keywords:
  10. Wiener Filter ; Image Registration ; Similarity Measure ; B-Spline Function ; Thin Plate Spline ; Magnetic Resonance Imagin (MRI)

 Digital Object List

 Bookmark

...see more