People - Research Staff / Muharrem Mercimek

Iterative and Linear Deconvolution Methods

The problem of deconvolution is fundamental to many problems of image restoration. Many of the practical modalities share the property that an imaging process distorts an object. The signal of interest is degraded by noise, blur and the presence of other extraneous data. The important data is hidden by noise and less important signals. In many cases the blurring is known but the problem is complicated by either noise that is due to low light levels or sensor readout noise. In this research several deconvolution algorithms are studied.

Nano Imaging

•In this research, a very unique data obtained with a high-resolution electron microscope, was studied.

•An iterative non-negative least squares solver, proposed by Dougherty, towards approximating the true image function is used. •Two main problems effecting the 3D visualization of the data are handled;

Misalignments of the depth sections.

–Degradations seen on the image model.

•A visually significant enhancement is obtained.

•

Thermal and Visual Image registration using MI

Face images of both thermal and visual images are used multi-modal image registration for image fusion purpose is investigated.

•

• Normalized mutual information value for heuristic parameter search step of is stored. Maximization of the MI gives us the appropriate alignment for registration. Due to its intensity based calculations in the different modalities this method tries to find similar regions. If the similar regions are overlapped MI of the overlapping part will be the best alignment, and registration.

•

•A smooth continuously increasing or decreasing MI function is hard to find. A general algorithm that does not trapped into local minima will be the solution of this problem. In the parameter search algorithm scaling, translation, and rotation are considered.

• •Face images are from Iris database acquired by face recognition and image fusion team. Pose change should be minimized while taking the photos of different people.

•

Image registration using Mutual information with image blending

For the evaluation of the registered images blending or warping is applied.

•

•Assuming the registration is done with small rotations (they are actually between 0 and 5 degrees) and illumination change is along one direction we applied a first degree linear blending algorithm along one direction. We super positioned floating and reference images after we applied blending to both of them.

•

After blending we had a smooth registered image.

3D reconstruction of automotive parts and terrain

To reconstruct the surfaces of small mechanical objects and terrains using primitive relationships and functions of 3D points acquired with laser scanners and to analyze 3D images. Genex 3D FaceCam scanner for small objects, Riegl Laser Scanner for terrain are used as data acquisition tools and for registration Rapidform04 software is used.

		Office:	209 Ferris Hall, 1508 Middle Drive University of Tennessee Knoxville, TN 37996-2100
		E-mail:	mmercime@utk.edu

	Office Hours	Meetings/Classes	Lunch Time
Monday	9.00-5.30	IRIS Lab Meeting 10:00	12:00-12:30
Tuesday	9.00-5.30		12:00-12:30
Wednesday	9.00-5.30		12:00-12:30
Thursday	9.00-5.30		12:00-12:30
Friday	9.00-5.30	IRIS Lab Meeting 10:30	12:00-12:30

		Iterative and Linear Deconvolution Methods The problem of deconvolution is fundamental to many problems of image restoration. Many of the practical modalities share the property that an imaging process distorts an object. The signal of interest is degraded by noise, blur and the presence of other extraneous data. The important data is hidden by noise and less important signals. In many cases the blurring is known but the problem is complicated by either noise that is due to low light levels or sensor readout noise. In this research several deconvolution algorithms are studied.

		Nano Imaging •In this research, a very unique data obtained with a high-resolution electron microscope, was studied. •An iterative non-negative least squares solver, proposed by Dougherty, towards approximating the true image function is used. •Two main problems effecting the 3D visualization of the data are handled; Misalignments of the depth sections. –Degradations seen on the image model. •A visually significant enhancement is obtained.

		• Thermal and Visual Image registration using MI Face images of both thermal and visual images are used multi-modal image registration for image fusion purpose is investigated. • • Normalized mutual information value for heuristic parameter search step of is stored. Maximization of the MI gives us the appropriate alignment for registration. Due to its intensity based calculations in the different modalities this method tries to find similar regions. If the similar regions are overlapped MI of the overlapping part will be the best alignment, and registration. • •A smooth continuously increasing or decreasing MI function is hard to find. A general algorithm that does not trapped into local minima will be the solution of this problem. In the parameter search algorithm scaling, translation, and rotation are considered. • •Face images are from Iris database acquired by face recognition and image fusion team. Pose change should be minimized while taking the photos of different people. •

		• Image registration using Mutual information with image blending For the evaluation of the registered images blending or warping is applied. • •Assuming the registration is done with small rotations (they are actually between 0 and 5 degrees) and illumination change is along one direction we applied a first degree linear blending algorithm along one direction. We super positioned floating and reference images after we applied blending to both of them. • After blending we had a smooth registered image.

		3D reconstruction of automotive parts and terrain To reconstruct the surfaces of small mechanical objects and terrains using primitive relationships and functions of 3D points acquired with laser scanners and to analyze 3D images. Genex 3D FaceCam scanner for small objects, Riegl Laser Scanner for terrain are used as data acquisition tools and for registration Rapidform04 software is used.