Next Generation CT (Computed Tomography) Student: Jaewon Kim Advisor: Ramesh Raskar Reader: V. Michael Bove Yasuhiro Mukaigawa Fredo Durand
Dream & Goal Dream Thesis Goal Understanding inside of a human body with portable system Wearable, always-on system Non scanning system and visible light for health and HCI (no x-ray CT) Thesis Prototype: 3D finger vein no x-ray CT reconstruction/human Id system
How will CT evolve? 5 th ? 1 st : Linear & rotational CT 2 nd : Multiple scanning CT Non-scanning No Synchronization Compact Safe 3 rd : Fan-beam CT 4 th : spiral CT or helical CT
Traditional CT Benefit See through human body with X-ray Limitation: Scanning mechanism makes CT large and cumbersome Expensive Exposure to X-rays Poor dynamic volume reconstruction Standard CT system
Thesis Goal A. Avoid mechanical scanning (Also applicable to actual X-ray CT system) Single shot CT of translucent (non-scattering) objects B. Visible light CT and overcome scattering Imaging through scattering objects
Goal A: Non scanning CT Prior work Shield Field Imaging (Lanman, Raskar, Agrawal and Gabriel[2008] ) Opaque object reconstruction from a single photo Limitation : Lacks support for translucent object A single shot photo Shield Field system 3D Reconstruction Decoupling to multi-view images
Goal A: Non scanning CT My contribution 3D Reconstruction of partially transparent objects with a single shot photo Limitation : low resolution due to decoded multi-view images single shot photo Single shot CT setup 3D Reconstruction Decoded multi-view images
Goal A: Non scanning CT My contribution Reconstruction of partially transparent objects using visible light Multi-view images(resolution is 100x200) 3D reconstruction results in high resolution
Goal B: Visible light CT Prior work Separation of Direct and Scattered component of reflection (Nayar, Krishnan, Grossberg and Raskar [2006] ) Reflection-based method using projected patterns Limitation : not single shot, no 3d reconstruction, no transmission mode = + Scattered component Direct Image Photo of mannequin and shower curtain
Goal B: Visible light CT My contribution Transmission-Mode Direct-Global Separation of Scattering Objects (will be submitted to CVPR 2010) a single LED and lenslet(pinhole) array method Acquiring a clear image for objects inside of scattering media + = Photo of mannequin Direct Image Scattered component and shower curtain Imaging Setup
Example Application Single shot 3D vein reconstruction Finger vein imaging using near-infrared LED by Hitachi and Sony vein authentication+lightfield camera Near-infrared image of finger vein (ML disclosure filed as 3D Finger Vein Authentication )
Limitations & Future directions Limitations Difficult to see through dense area of a human body with visible light Decreased resolution of processed light field images Need for multiple light sources for non scanning CT system Future Replacing traditional CT system by single shot compact system Health and HCI applications with portable and safe CT machines Self imaging of CT images by patients
Evaluation Goal A: dynamic volume reconstruction of translucent objects with a resolution of 100x200x200 Evaluation: Comparing 3D reconstruction results in proposed and normal tomographic techniques Comparing SNR between lightfield and normal images Goal B: 3D finger vein CT system with a resolution of 100x100x100 Evaluation: Id using vein pattern from 20 people database by 3D matching
Dream & Goal Dream Thesis Goal Understanding inside of a human body with portable system Wearable, always-on system Non scanning system and visible light for health and HCI (no x-ray CT) Thesis Prototype: 3D finger vein no x-ray CT reconstruction/human Id system
Recommend
More recommend