000			02841nam a22001577a 4500
082			_a621.382,RIA
100			_aRiaz, Muhammad Mohsin _935188
245			_aThrough Wall Image Enhancement / _cMuhammad Mohsin Riaz
260			_aRawalpindi, _bMCS (NUST), _c2013
300			_axxii,100 -p,
505			_aThrough wall imaging has wide range of civil and military applications (like attaining hostage or terrorist information, weapons, explosive materials, building structures, detection buried people etc). Through wall imaging process involve a wall as an obstacle, which causes strong wall reflections which are termed as clutter. Moreover, due to reflections from surrounding and some other artifacts (like system and false targets etc) noise components are also present in the received signals. To mitigate, the clutters and noise a robust pre-processing (image enhancement) step is essential. A review of state of the art through wall imaging along with their limitations and drawbacks is presented. Various statistical methods based improvements are proposed for through wall image enhancement. The target subspace estimation concept is used for through wall image enhancement. The input image is decomposed into different spectral components using singular value decomposition and then different techniques (conventional, clustering and information theoretic criterions) are applied to extract target spectral components. Minimum description length criterion based technique is proposed which successfully estimate target spectral components and extract single and multiple targets more accurately. Often in through wall imaging, the boundaries of clutter noise and targets are not sharpey defined which limit the performance of subspace estimation schemes. This motivates the use of weighting based addition of different spectral components for improved results. Singular values and their differences are used in combination with fuzzy logic to calculate weights. K-means clustering based adjustment of membership functions is proposed. Mamdani and Takagi-Sugeno inference engines are used for input/output mapping. Later textural features extracted from spectral components are combined to get further improvements in results. Beside singular value decomposition based schemes, other statistical methods for through wall image enhancement are explored. A brief discussion on different statistical methods which can be combined with proposed techniques is presented. The applications of proposed fuzzy logic based weight computation in different domains including ground penetrating radar, medical imaging and image fusion are also explored.
650			_aPhD Electrical Engineering Thesis _9133107
651			_aPhD EE Thesis _9133108
700			_aSupervised by Dr. Abdul Ghafoor _9132894
942			_2ddc _cTHE
999			_c608484 _d608484