Yeong-Hwa Kim

Image feature and noise detection based on statistical hypothesis tests and their applications in noise reduction

In many video processing applications in the field of consumer electronics such as digital TV, it is well understood that the presence of a noise limits the performance of video enhancement functions due to the time-varying characteristics of the noise. The basic difficulty is that the noise and the signal are difficult to be distinguished. This paper proposes image feature and noise detection algorithms, which effectively distinguish the noise from the image feature or vice versa. Specifically, the proposed algorithms provide a way of measuring the degree of noise with respect to the degree of image feature. The fundamental idea behind the proposed algorithms is to derive a statistical measure to estimate the fact that a noise has a random characteristic whereas an image feature has a spatial correlation among the associated neighbor samples. With the proposed algorithms, many video enhancement algorithms such as noise reduction or sharpness enhancement can be adaptively performed although a time varying noise is presented.

The Behrens-Fisher problem revisited: a Bayes-frequentist synthesis

The Behrens-Fisher problem concerns the inference for the difference between the means of two normal populations whose ratio of variances is unknown. In this situation, Fishers fiducial interval differs markedly from the Neyman-Pearson confidence interval. A prior proposed by Jeffreys leads to a credible interval that is equivalent to Fishers solution but it carries a different interpretation. The authors propose an alternative prior leading to a credible interval whose asymptotic coverage probability matches the fre- quentist coverage probability more accurately than the interval of Jeffreys. Their simulation results indicate excellent matching even in small samples.

Adaptive noise reduction algorithms based on statistical hypotheses tests

In many video processing applications, the presence of a random noise is troublesome since most video enhancement functions produce visual artifacts if a priori of the noise is incorrect. The basic difficulty is that the noise and the signal are difficult to be distinguished. It was shown that the noise and image feature detection problem can be converted to statistical hypotheses tests based on the sample correlation in different orientations. In this paper, to further elaborate these hypotheses, we propose parametric, semi- parametric, and nonparametric statistical tests by combining with adaptive median filters. The proposed algorithms provide ways of measuring the degree of noise with respect to the degree of image feature, and the proposed adaptive noise reduction filtering framework provides good performance when the underlying noises are from Gaussian or non-Gaussian distributions. Simulation results for noise reduction show that the Bartlett and the Levene tests perform better regardless of the noise characteristics. Applications of the proposed algorithms can be found in digital TV, camcorders, digital cameras, and DVD players.

Feature and noise adaptive unsharp masking based on statistical hypotheses test

The conventional unsharp masking (UM) enhances the visual appearances of images by adding their amplified high frequency components. However, the noise component of the input image also tends to be amplified due to the nature of the UM. Hence, the application of the conventional UM is not suitable when noise is present. This paper exploits the statistical theories proposed in A. Polesel, et al., (1997) and Y.-H. Kim and J. Lee, (Nov 2005) for detecting noise and image feature of the input image so that the UM could be adaptively applied accordingly. By applying the proposed algorithm, it is made possible to enhance local contrast of the image, especially, the area with small details, without boosting up the noise counterpart. This results in natural looking output image.

Dynamic range compression and contrast enhancement for digital images in the compressed domain

We develop a simple and efficient algorithm for dynamic range compression and contrast enhancement of digital images in the compressed domain. The basic idea of our approach is to separate illumination and reflectance components of an image in the compressed domain. We adjust the amount of contribution of the illumination component to effectively compress the dynamic range of the image. For contrast enhancement, we modify the reflectance component based on a new measure of the spectral contents of the image. The spectral content measure is computed from the energy distribution across different spectral bands in a discrete cosine transform (DCT) block. The advantages of the proposed algorithm are (1) high dynamic range scenes are effectively mapped to the smaller dynamic range of the image, (2) the details in very dark or bright areas become clearly visible, (3) the computational cost is low, and (4) the compressibility of the original image is not affected by the algorithm. We evaluate the performance of the proposed algorithm with well-known existing methods, such as histogram equalization and -rooting algorithm, using a few different enhancement quality metrics.

Adaptive Noise Reduction Algorithm for an Image Based on a Bayesian Method

Noise reduction is an important issue in the field of image processing because image noise lowers the quality of the original pure image. The basic difficulty is that the noise and the signal are not easily distinguished. Simple smoothing is the most basic and important procedure to effectively remove the noise; however, the weakness is that the feature area is simultaneously blurred. In this research, we use ways to measure the degree of noise with respect to the degree of image features and propose a Bayesian noise reduction method based on MAP (maximum a posteriori). Simulation results show that the proposed adaptive noise reduction algorithm using Bayesian MAP provides good performance regardless of the level of noise variance.

Posterior Inference in Single-Index Models

A single-index model is useful in fields which employ multidimensional regression models. Many methods have been developed in parametric and nonparametric approaches. In this paper, posterior inference is considered and a wavelet series is thought of as a function approximated to a true function in the single-index model. The posterior inference needs a prior distribution for each parameter estimated. A prior distribution of each coefficient of the wavelet series is proposed as a hierarchical distribution. A direction is assumed with a unit vector and affects estimate of the true function. Because of the constraint of the direction, a transformation, a spherical polar coordinate , of the direction is required. Since the posterior distribution of the direction is unknown, we apply a Metropolis-Hastings algorithm to generate random samples of the direction. Through a Monte Carlo simulation we investigate estimates of the true function and the direction.