Yeong-Taeg Kim

Contrast enhancement using brightness preserving bi-histogram equalization

Histogram equalization is widely used for contrast enhancement in a variety of applications due to its simple function and effectiveness. Examples include medical image processing and radar signal processing. One drawback of the histogram equalization can be found on the fact that the brightness of an image can be changed after the histogram equalization, which is mainly due to the flattening property of the histogram equalization. Thus, it is rarely utilized in consumer electronic products such as TV where preserving the original input brightness may be necessary in order not to introduce unnecessary visual deterioration. This paper proposes a novel extension of histogram equalization to overcome such a drawback of histogram equalization. The essence of the proposed algorithm is to utilize independent histogram equalizations separately over two subimages obtained by decomposing the input image based on its mean with a constraint that the resulting equalized subimages are bounded by each other around the input mean. It is shown mathematically that the proposed algorithm preserves the mean brightness of a given image significantly well compared to typical histogram equalization while enhancing the contrast and, thus, provides a natural enhancement that can be utilized in consumer electronic products.

Deinterlacing algorithm based on sparse wide-vector correlations

In this paper, we propose a new deinterlacing algorithm based on sparse wide vector correlations, which is an extension of the deinterlacing algorithm previously proposed by the author, aimed to reduce the H/W complexity in applications. The proposed algorithm is developed mainly for the format conversion problem encountered in current HDTV systems, but can also be applicable to the double rate conversion problem in the NTSC system. By exploiting the edge oriented spatial interpolation based on the wide vector correlations, visually annoying artifacts caused by interlacing such as a serrate line, line crawling, a line flicker, and a large area flicker can be remarkably reduced since the use of the wide vector correlation increases the range of the orientations that can be detected, and by introducing sparse vectors the H/W complexity for realizing the algorithm in applications can be significantly reduced. Simulations are also provided indicating that the proposed algorithm results in a high performance comparable to the performance of the deinterlacing algorithm based on wide vector correlations.

Quantized bi-histogram equalization

Histogram equalization is a widely used scheme for contrast enhancement in a variety of applications due to its simple function and effectiveness. One possible drawback of the histogram equalization is that it can change the mean brightness of an image significantly as a consequence of histogram flattening. Clearly, this is not a desirable property when preserving the original mean brightness of a given image is necessary. As an effort to overcome such drawback for extending the applications of the histogram equalization in consumer electronic products, bi-histogram equalization has been proposed by the author which is capable of preserving the mean brightness of an image while it performs contrast enhancement. The essence of the bi-histogram equalization is to utilize independent histogram equalizations separately over two subimages obtained by decomposing the input image based on its mean. A simplified version of the bi-histogram equalization is proposed, which is referred to as the quantized bi-histogram equalization. The proposed algorithm provides a much simpler hardware (H/W) structure than the bi-histogram equalization since it is based on the cumulative density function of a quantized image. Thus, the realization of bi-histogram equalization in H/W is feasible, which leads to versatile applications in the field of consumer electronics.

An adaptive color transient improvement algorithm

In color TV broadcasting standards, such as NTSC and PAL, the limited bandwidth of the chrominance signals produces relatively slow chrominance transitions, causing smeared color edges in the received images. In this paper, we present an adaptive color transient improvement algorithm for enhancing color edge transition of images. By using a local image feature dependent gain control function to adaptively control the enhancement of the color transient, the proposed algorithm produces steeper natural color edge transitions and prevents the overshoot effects and the amplification of the noise.

Low line memory visually lossless compression for color images using non-uniform quantizers

The paper proposes a novel method to compress color images with imperceptible quality loss. The algorithm explores the difference in error perceptibility of human visual system (HVS) for various areas. It is done by implementing different non-uniform quantizers for flat, detail and random blocks of pixels. These blocks are classified based on principle component analysis (PCA) and prediction error. For hardware implementation purpose, the algorithm is designed to use very low line memory. Simulation results show that the proposed compression is visually lossless in all categories of tested images with high compression ratio.

Motion decision feedback deinterlacing algorithms

As an effort to improve the video quality of a digital TV (HDTV) system, a robust motion adaptive deinterlacing algorithms based on motion decision feedback rules are proposed, which, in principle, utilizes the motion decision information of the past and the current pictures. With the proposed algorithm, visual artifacts due to the failure in detecting a fast repetitive motion embedded in an interlaced video sequence can be significantly reduced. Hence, with the proposed algorithms, much more pleasing video quality can be realized for the HDTV systems than typical motion adaptive deinterlacing algorithms.

An efficient neural network based indoor-outdoor scene classification algorithm

This paper presents an efficient indoor-outdoor scene classification algorithm based on neural network learning. To facilitate the hardware implementation, efficient image (composite) feature extraction based on image color, entropy, DCT coefficients, and edge orientation was developed. The trained scene classifiers demonstrate robust performance.

An image sharpening algorithm for high magnification image zooming

Traditional image sharpening methods usually introduce over shootings along the edges, and do not perform well for sharpening an image resulting from high magnification zooming. In this paper, we present an image sharpening algorithm which is overshooting free and particularly suitable for high magnification zooming situation. Our algorithm modifies the pixel value based on the desired changes for its gradient profile, and the sharpening parameters can be learned for the particular zooming rate. Experimental results show that the proposed algorithm outperforms the conventional image sharpening methods.

A simple and effective algorithm for false contour reduction in digital television

False contours are visually unpleasant stage-like edges within areas which are supposed to be smooth. Smoothing of these false contours usually results in losing some small magnitude details. In this paper, we present a false contour reduction algorithm and its hardware design for digital televisions. Although the hardware design has a limitation of two line memory, the proposed method performs reasonably well in reducing the false contour and preserving small magnitude details. Experimental results show that the proposed method has better performance.

Decolored Ll filter lattice: a subclass of Ll filters useful in practice

The Ll filters constitute a wide class of filters including both linear and L filters, and it has been shown that they outperform linear and L filters. A close examination of the optimization of the Ll filters in real applications, however, casts a problem of importance that the Ll filters can be degraded from insufficient signal statistics for optimization since a sufficiently long training data sequence is not tractable in practice for the Ll filters, especially, when the the number of input taps is large. We formally propose a new class of order-statistic filters referred to as the decolored Ll filter lattice, which exploits a decolored rank permutation space. By decoloring the rank permutation space, having sufficient training data for optimization in practice is feasible. Various filters are obtained in the class of decolored Ll filters for different orders of decoloring. It is shown that the set of decolored Ll filters of all orders form a lattice structure. At the simplest level in the structure, the decolored Ll filters reduce to a simple linear (FIR) filter, and we obtain the Ll filter at the highest level in the structure.