19-03-2012, 12:42 PM
SUPER-RESOLUTION RECONSTRUCTION OF HYPERSPECTRAL IMAGES
ABSTRACT
SUPER-RESOLUTION RECONSTRUCTION OF HYPERSPECTRAL IMAGES
Esmy Thekketh
Nowadays medical diagnosis is principally supported by the imaging techniques. Several imaging modalities such as magnetic resonance imaging (MRI), computed tomography (CT), ultrasonography, Doppler scanning, and nuclear imaging have completely expanded medical imaging field. Recently Hyperspectral imaging (HSI), has emerged as a new member of the family of the medical imaging modalities. HSI provides a powerful tool for non-invasive tissue analyses. This technology is able to capture both the spatial and spectral data of an organ or tissue in one snapshot. In other words, the imaging system produces many narrow band images at different wavelengths. HSI can visualize invisible wavelength regions and bring them to the human vision range. In this paper, I introduce a novel super-resolution method for hyperspectral images. An integral part of our work is to model the hyperspectral image acquisition process. We propose a model that enables us to represent the hyperspectral observations from different wavelengths as weighted linear combinations of a small number of basis image planes. Then, a method for applying super resolution to hyperspectral images using this model is presented. The method fuses information from multiple observations and spectral bands to improve spatial resolution and reconstruct the spectrum of the observed scene as a combination of a small number of spectral basis functions.
CHAPTER 1
INTRODUCTION
Nowadays medical diagnosis is principally supported by the imaging techniques. Several imaging modalities have completely expanded medical imaging field. Recently Hyperspectral imaging, has emerged as a new member of the family of the medical imaging modalities. It provides a powerful tool for non-invasive tissues analyses. This technology is able to capture both the spatial and spectral data of an organ or tissue in one snapshot. This system captures full neighboring spectral data with spectral and spatial information. HSI can visualize invisible wavelength regions and bring them to the human vision range. Since the spatial resolution is a key parameter in many applications related to space imagery, it is obvious that any improvement here is important. To improve the spatial resolution of Hyperspectral images, we can make use of super-resolution techniques together with the information at different wavelengths of the sense. Any effective super-resolution method requires frequency aliasing to be present in the low-resolution observation images.
In this seminar I propose a novel Hyperspectral image acquisition model that enables us to represent Hyperspectral observations from different wavelengths as weighted linear combinations of a small number of aliased and blurred basis image planes whose pixel values correspond to the principle component coefficients. We proceed by formulating the reconstruction process as the inverse problem of finding a hi-res target Hyperspectral image that agrees best with the observations under the proposed model. Then, a set-theoretic method is used to solve the inverse problem.
CHAPTER 2
HYPERSPECTRAL IMAGING
Hyperspectral captures and analyses data from across the electromagnetic spectrum. This technology extends the human vision that just sees visible light. Hyperspectral imaging can visualize the visible light as well as near-infrared to infrared. The difference between Hyperspectral and multi-spectral imaging is usually defined according to the number of spectral bands. Multi-spectral image contains tens of bands. However, Hyperspectral image contains hundreds to thousands of bands. Hyperspectral images are captured by one sensor that captures a set of contiguous bands. However, multi- spectral is a set of spectral bands that are not typically contiguous and can be captured by multiple sensors.
The spectral graph of the average spectrum from the pig’s peritoneum, spleen, colon, and urinary bladder are shown in the four graphs. The graph depicts the reflectance for each wavelength in that region.
Hyperspectral sensors generate a two-dimensional spatial image along a third spectral dimension. Each pixel in the Hyperspectral image has a sequence of reflectance in different spectral wavelengths that can display the spectral signature of that pixel.
2.1 HYPERSPECTRAL SENSORS
The Hyperspectral sensors are instruments for capturing many images in different adjacent wavelengths of an illuminated region corresponding to the entrance slit. The main components of a Hyperspectral camera are shown in Figure As light sources, two halogen lamps illuminate the object to be captured. The camera’s objective lens collects the radiation from the object and projects an image on the entrance slit plane. The slit determines the instantaneous field of imaging in spatial directions. The radiation from the slit is projected to the prism-grating-prism (PGP) components. Therefore, the propagation angle of the radiation changes depending on its wavelength. Then it is focused on the matrix detector. Every object’s point is represented on the matrix detector by a series of monochromatic points that makes a continuous spectrum in the direction of the spectral axis.
Whish-broom scanning, filtered imaging and push-broom are three main designs for Hyperspectral cameras. Whish-broom scanner captures the spectral dimension pixel-to-pixel. Filtered camera captures two spatial dimensions and temporally samples the spectral dimension. The capturing technique of ImSpector sensors is a push-broom scanning. In this type of imaging spectrometer, the entrance slit limits the imaging field. The 2D detector matrix instantaneously captures the spectral dimension and one spatial dimension. The second spatial dimension is generated by scanning the object. By moving the camera’s field of view relative to the object, the second spatial dimension is created.
The PGP is composed of a special grating, two prisms, and an aperture stop. The special grating is located between two prisms and the aperture stop is set in contact with the grating. Short and long pass filters are usually placed between the grating and the prisms, eliminating unwanted wavebands and changing the spectral response. In this technique, since the filters are incorporated inside a PGP, the reflections from their surfaces can be eliminated. Unlike a direct vision prism where the dispersion is a non-linear small dispersion, the diffraction grating in the PGP supplies a large linear dispersion. All the wavelengths will be passed for only the small region of the object that is exactly in front of the entrance slit. By shifting the sensor between subsequent images, ultimately all parts of the object and all corresponding wavelengths are captured. Therefore, for each wavelength, a monochromatic spectral image can be constructed from the Hyperspectral image set.
2.2 CAPTURING SETUP
The ImSpector sensors capture the images using the push-broom scanning technique. Therefore, to generate the second spatial dimension the object must be scanned i.e. the second spatial dimension is captured by moving the camera’s field of view relative to the object. The linear actuator, a ROBO Cylinder Slider, is used to move the camera. This actuator is controlled by a controller. The actuator is connected to the controller by two cables: the encoder cable and the motor cable. The movement and velocity are adjusted by a setting tool that is connected to the controller. The actuator moves the camera with a constant velocity.
The acquisition setup consists of a pair of 500 N halogen lamps with diffusing reflectors as the light sources and the computer-controlled linear actuator. The linear actuator is fixed on a bridge installed over the surgical bed and the camera has been calibrated and fixed on the frame. Therefore, the distance between the lens and the abdomen is constant and a fairly uniform illumination of the subject is provided by using the two halogen lamps.
2.3 DATA NORMALIZATION
The captured data should be normalized to treat the spectral non-uniformity of the illumination device. The raw data are changed by illumination conditions and temperature. Therefore, the radiance data were normalized to yield the radiance of the specimen.White reference and dark current are two data that should be captured for normalization.White reference is the spectrum acquired by the Hyperspectral sensor corresponding to the white reference and dark current is a dark image acquired by the system in the absence of light. White reference is used to show the maximum reflectance in each wavelength. Dark current spectroscopy is used to address the defects by calculating the peaks in the dark current spectrum with temperature. To perform this pre-processing step, the radiance of a standard reference white board placed in the scene and the dark current are measured by keeping the camera shutter closed. Then the raw data are corrected to reflectance using the following equation: