Kangeui Lee

Statistical reconstruction using dual formulation of subband-wise total variation regularization (SDST) for limited angle tomography

In this paper, a novel reconstruction algorithm for limited angle tomography using total variation (TV) regularization is presented. Inspired by duality-based TV minimization in denoising and deblurring applications, we derived a TV regularized statistical reconstruction algorithm composed of relatively simple and structured operations such as discrete gradient and divergence calculations, which presents an effective way to introduce TV regularization to the statistical reconstruction. In initial tests with real data from a digital breast tomosynthesis system, the proposed algorithm showed reliable reconstructions for low dose conditions.

A digital ISO expansion technique for digital cameras

Markets demands of digital cameras for higher sensitivity capability under low-light conditions are remarkably increasing nowadays. The digital camera market is now a tough race for providing higher ISO capability. In this paper, we explore an approach for increasing maximum ISO capability of digital cameras without changing any structure of an image sensor or CFA. Our method is directly applied to the raw Bayer pattern CFA image to avoid non-linearity characteristics and noise amplification which are usually deteriorated after ISP (Image Signal Processor) of digital cameras. The proposed method fuses multiple short exposed images which are noisy, but less blurred. Our approach is designed to avoid the ghost artifact caused by hand-shaking and object motion. In order to achieve a desired ISO image quality, both low frequency chromatic noise and fine-grain noise that usually appear in high ISO images are removed and then we modify the different layers which are created by a two-scale non-linear decomposition of an image. Once our approach is performed on an input Bayer pattern CFA image, the resultant Bayer image is further processed by ISP to obtain a fully processed RGB image. The performance of our proposed approach is evaluated by comparing SNR (Signal to Noise Ratio), MTF50 (Modulation Transfer Function), color error ∝E*ab and visual quality with reference images whose exposure times are properly extended into a variety of target sensitivity.

Exposure-adaptive Color Image Enhancement

In many cases, it is not possible to faithfully capture shadow and highlight image data of a high dynamic range (HDR) scene using a common digital camera, due to its narrow dynamic range (DR). Conventional solutions tried to solve the problem with an captured image which has saturated highlight and/or lack of shadow information. In this situation, we introduce a color image enhancing method with the scene-adaptive exposure control. First, our method recommends an optimal exposure to obtain more information in highlight by the histogram-based scene analysis. Next, the proposed luminance and contrast enhancement is performed on the captured image. The main processing consists of luminance enhancement, multi-band contrast stretching, and color compensation. The luminance and chrominance components of input RGB data is separated by converting into HSV color space. The luminance is increased using an adaptive log function. Multi-band contrast stretching functions are applied to each sub-band to enhance shadow and highlight at the same time. To remove boundary discontinuities between sub-bands, the multi-level low-pass filtering is employed. The blurred image data represents local illumination while the contrast-stretched details correspond to reflectance of the scene. The restored luminance image is produced by the combination of multi-band contrast stretched image and multilevel low-pass filtered image. Color compensation proportional to the amount of luminance enhancement is applied to make an output image.

Information theoretic discrepancy based iterative reconstruction for transmission tomography

The monochromatic approximation, which postulates that a source emits monochromatic radiation, has been widely used in the transmission tomography. However, due to the ignorance of energy dependency, image degradations such as beam hardening artifact are often occurred. In this paper, we present novel reconstruction algorithms to reflect the exact polychromatic model. Departing from the conventional algebraic and statistical reconstructions, the generalized information theoretic discrepancy (GID) is employed as the new data fidelity metric. By using the particular features of the GID, the cost function is derived in terms of imaginary variables, which incorporates energy dependency and leads to a tractable optimization problem even without the monochromatic approximation. In preliminary experiments with a simulated dual energy CT and a real experimental tomosyn-thesis, the proposed information theoretic discrepancy based iterative reconstruction (IDIR) algorithm showed superior performances over conventional reconstruction schemes.

Regularized polychromatic reconstruction for transmission tomography

A polychromatic reconstruction algorithm that accounts for the exact physical model of transmission tomography is presented. Based on the equivalence between the Poisson log-likelihood function and the I-divergence, we derived a fast convergencing algorithm with a pixel-wise updating scheme, which is an extended version of the AM-ICD algorithm. The objective function in each iteration consists of approximated I-divergence and the generalized Gaussian Markov random field (GGMRF) model based regularization term for preventing diverging due to additive noise and the approximation of I-divergence. In a simulation study, we observed that the beam hardening artifact was significantly reduced in the extended AM-ICD algorithm with the use of a reasonable number of iterations. In addition, the proposed algorithm also showed reliable reconstruction results even for low dose conditions.

Limited Data Tomographic Image Reconstruction via Dual Formulation of Total Variation Minimization

The X-ray mammography is the primary imaging modality for breast cancer screening. For the dense breast, however, the mammogram is usually difficult to read due to tissue overlap problem caused by the superposition of normal tissues. The digital breast tomosynthesis (DBT) that measures several low dose projections over a limited angle range may be an alternative modality for breast imaging, since it allows the visualization of the cross-sectional information of breast. The DBT, however, may suffer from the aliasing artifact and the severe noise corruption. To overcome these problems, a total variation (TV) regularized statistical reconstruction algorithm is presented. Inspired by the dual formulation of TV minimization in denoising and deblurring problems, we derived a gradient-type algorithm based on statistical model of X-ray tomography. The objective function is comprised of a data fidelity term derived from the statistical model and a TV regularization term. The gradient of the objective function can be easily calculated using simple operations in terms of auxiliary variables. After a descending step, the data fidelity term is renewed in each iteration. Since the proposed algorithm can be implemented without sophisticated operations such as matrix inverse, it provides an efficient way to include the TV regularization in the statistical reconstruction method, which results in a fast and robust estimation for low dose projections over the limited angle range. Initial tests with an experimental DBT system confirmed our finding.