Loading...

Multispectral Fuzzy Image Segmentation

Hasanzadeh, Maryam | 2009

868 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 39973 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Kasaei, Shohreh
  7. Abstract:
  8. Image segmentation is middle and an important task in image analysis and machine vision applications. The output images of imaging systems are often fuzzy because of noise, limitation in spatial and temporal resolution, blurring and intensity inhomogeneity in the objects. The goal of this thesis is exploring the fuzzy methods in multispectral image segmentation and proposing a new one to solve some of the recent difficulties and problems. The difficulties and problems such as simultaneous utilization of spatial and spectral information, necessity for dimension reduction, spatial and spectral and intra-cluster image information redundancy, existence of regions with widely varying size, existence of noise while important details are also exist and the required high accuracy in applications are considered in this thesis. Regarding the above mentioned problems, the proposed method is based on the combination of two well-known fuzzy methods; fuzzy clustering and fuzzy connectedness. The former does not inherently consider the spatial relation among image pixels and the later is focused on spatial relation the most. In the proposed method the simultaneous utilization of spatial and spectral fuzzy information is possible so the segmentation accuracy will be improved especially in noisy images with important details. For fuzzy clustering step a modified version of fuzzy C-means which is capable in detection of small regions is proposed called size-weighted FCM. The main idea in this method is removing the intra-cluster redundancy and emphasizing on small clusters. Moreover the multispectral watershed transform is utilized before clustering step to remove the spatial redundancy and improves the accuracy and efficiency of segmentation method. The remained problem is the method of utilization of different bands of the multispectral image which depends on the kind of the figures and the utilized method (supervised and unsupervised). We have evaluated the proposed method based on the two kinds of multispectral images. The first one is the multispectral brain magnetic resonance images in which all of the spectral bands are used. The second one is the multispectral remote sensing images in which the dimension reduction process is applied. More over a the propose method is applied on a created grayscale synthetic dataset for better evaluation. The performance analysis shows the ability of the method in accurate segmentation and detection of small regions when comparing with other recent methods.

  9. Keywords:
  10. Image ; Watershed Method ; Image Segmentation ; Remote Sensing ; Fuzzy Clustering ; Fuzzy Connectedness ; Multispectral Image

 Digital Object List

 Bookmark