Sung-su Kim

Multiple object decomposition based on independent component analysis of multi-energy x-ray projections

X-ray projection is not effective for representing complex overlapping objects. This paper presents a novel computational framework to decompose X-ray projections into multiple images with non-overlapping objects that are differentiated by their own material compositions. Based on energy-dependent X-ray attenuation characteristics for each material, multiple energy X-ray images are analyzed to obtain material-selective images, which correspond to projections of basis materials that constitute objects. We show that material-selective images can be considered as linear mixtures of independent components that are associated with object-selective images. As a result, multiple objects can be decomposed by independent component analysis (ICA) of material-selective images or ICA of multiple monochromatic energy X-ray images. To demonstrate the concept of the proposed method, we apply it to simulated images based on a 3-D human model.

Fast and accurate auto focusing algorithm based on two defocused images using discrete cosine transform

This paper describes the new method for fast auto focusing in image capturing devices. This is achieved by using two defocused images. At two prefixed lens positions, two defocused images are taken and defocused blur levels in each image are estimated using Discrete Cosine Transform (DCT). These DCT values can be classified into distance from the image capturing device to main object, so we can make distance vs. defocused blur level classifier. With this classifier, relation between two defocused blur levels can give the device the best focused lens step. In the case of ordinary auto focusing like Depth from Focus (DFF), it needs several defocused images and compares high frequency components in each image. Also known as hill-climbing method, the process requires about half number of images in all focus lens steps for focusing in general. Since this new method requires only two defocused images, it can save lots of time for focusing or reduce shutter lag time. Compared to existing Depth from Defocus (DFD) which uses two defocused images, this new algorithm is simple and accurate as DFF method. Because of this simplicity and accuracy, this method can also be applied to fast 3D depth map construction.

A novel material decomposition algorithm for multienergy X-ray radiography systems

A novel algorithm for multienergy X-ray radiography systems that simultaneously estimate multiple measurements with energy diversity is presented. In contrast to conventional dual source X-ray radiography which utilizes simple weighted subtractions, the proposed algorithm aims for an accurate material decomposition of more than two materials. The numerical simulation as well as the real experiment using an energy discriminating detector confirmed our finding.

A new calibration method and tissue cancellation in dual energy mammography

Mammography is stil the major tool for screening and diagnostic examinations of early breast cancer detection. It gives opportunities for early detection of breast cancer. For early detection of breast cancer, the mass visuality of mammography is very important. Direct mapping is suggested through which the breast radiography relates to the reference phantom radiography by finding the closest intensity values between those radiography images at both energy. With the proposed method, we could enhance the visuality of mass lesions in breast. And the results show that the proposed method has good performance, showing the possibility of cancer detection mammography.

Mammogram enhancement using multi-energy x-ray

This paper proposes a new method to improve contrast of a mammogram using multi-energy x-ray (MEX) images. The x-ray attenuation differences among breast tissues increase as incident photons have lower energy. Thus an image obtained by a narrow low energy spectrum has higher contrast than a full (wide) energy spectrum image. The proposed mammogram enhancement utilizes this fact using MEX images. Lowpass data of a low energy spectrum image and high frequency components of a wide energy spectrum image are combined to have high contrast and low noise. Nonsubsampled contourlet transform (NSCT) is employed to decompose image data into multi-scale and multidirectional information. The NSCT overcomes the shortage of directions of wavelet transform by expressing smoothness along contours sufficiently. The outcome of the transform is a lowpass subband and multiple bandpass directional subbands. First, the lowpass subband coefficients of a wide energy spectrum image are substituted by those of a low energy spectrum image. Before the coefficient modification, the low energy spectrum image is processed to have high contrast and sharp details. Next, for the bandpass directional subbands, the locally adaptive bivariate shrinkage of contourlet coefficients is applied to suppress noise. The bivariate shrinkage function exploits interscale dependency of coefficients. Local contrast of the resultant mammogram is considerably enhanced and shows clear fibroglandular tissue structures. Experimental results illustrate that the proposed method produces a high contrast and low noise level image, as compared to the conventional mammography based on a single energy spectrum image.

High contrast soft tissue imaging based on multi-energy x-ray

Breast soft tissues have similar x-ray attenuations to mass tissue. Overlapping breast tissue structure often obscures mass and microcalcification, essential to the early detection of breast cancer. In this paper, we propose new method to generate the high contrast mammogram with distinctive features of a breast cancer by using multiple images with different x-ray energy spectra. On the experiments with mammography simulation and real breast tissues, the proposed method has provided noticeable images with obvious mass structure and microcalifications.

Color balancing based upon gamut and temporal correlations

Image acquisition devices inherently do not have color constancy mechanism like human visual system. Machine color constancy problem can be circumvented using a white balancing technique based upon accurate illumination estimation. Unfortunately, previous study can give satisfactory results for both accuracy and stability under various conditions. To overcome these problems, we suggest a new method: spatial and temporal illumination estimation. This method, an evolution of the Retinex and Color by Correlation method, predicts on initial illuminant point, and estimates scene-illumination between the point and sub-gamuts derived by from luminance levels. The method proposed can raise estimation probability by not only detecting motion of scene reflectance but also by finding valid scenes using different information from sequential scenes. This proposed method outperforms recently developed algorithms.

Compensation method for color defects in PDP due to different time responses of phosphors

On a plasma display panel (PDP), luminous elements of red, green, and blue have different time responses. Therefore, a colored trails and edges appear behind and in front of moving objects. In order to reduce the color artifacts, this paper proposes a motion-based discoloring method. Discoloring values are modeled as linear functions of a motion vector to reduce hardware complexity. Experimental results show that the proposed method has effectively removed the colored trails and edges of moving objects. Moreover, the clear image sequences have been observed compared to the conventional ones.