Hagyong Han

A Space-Variant Luminance Map based Color Image Enhancement

Improvement of image quality has been highly demanding for users in digital imaging systems. A variety of image enhancement algorithms has been developed to control image enhancement factors. This paper presents a color image enhancement method that makes use of a space-variant luminance map (SVLM) for the local brightness characterization. Two-dimensional gamma correction combined with the SVLM is developed to increase the intensity at the dark region and vice versa at the bright region in the luminance domain. The enhanced luminance information is applied to the adaptive contrast enhancing process and the linear color restoration. The proposed algorithm reveals details of the input image as well as minimizes loss of the edge sharpness in the non-uniform and low illumination conditions. The developed enhancement process resulted in the lowest absolute mean brightness error (AMBE) in a low-complexity fashion. A series of experimental results are discussed in terms of visual quality, the enhancement performance, and the complexity.

New video enhancement preprocessor using the region-of-interest for the videoconferencing

The development of audio/video compression systems has made feasible digital multimedia applications. Under limited bandwidth environments, the human visual attention using region-of-interest (ROI) has been reported to improve the performance in the video coding. In this paper, we propose a ROI based video preprocessor method that deals with the perceptual quality in a low-bit rate communication environment. Three separated processes of the ROI detection, the image enhancement, and the boundary reduction are proposed to deliver better video quality at the videoconferencing application for use in a fixed camera and to be compatible as a preprocessor with the conventional video coding standards. A series of the simulation results and the subjective quality comparison assessment are discussed to show the quality improvement.

Color image enhancement method using a space-variant luminance map

This paper presents a method for improving the contrast of color images. A variable and nonlinear gamma characteristic is applied to the input image, depending on the local dependency of the intensity. In this work, we present a local contrast correction algorithm that allows for simultaneous low-light region and high-light region adjustments, based on simple pixel-wise gamma correction. Also, to minimize the change in color saturation between the input and the output image, we present the applied local characteristic.

An Effective Noise Estimator for Use in Noise Reduction

Conventional noise reduction filtering schemes realize limited improvements of the peak signal-to-noise ratio (PSNR) in the low-level noisy images. The flatness degree and the edge information are effectively used to estimate the noise volume. We propose a noise estimator for reducing noise in the AWGN (additive white gaussian noise) corrupted images using three intermediate image maps (FGM(flatness gray map), FIM(flatness index map), NEM(noise estimate map)). The proposed noise estimator is fed into the conventional noise reduction filters as a pre-processor. The performance of noise reduction is tested in the various AWGN corrupted images.

Face Region Tracking Improvement and Hardware Implementation for AF(Auto Focusing) Using Face to ROI

In this paper, we proposed a method about improving face tracking efficiency of face detection for AF system using the faces to the ROI. The conventional face detection system detecting faces based skin color uses the ratio of skin pixels of the present frame to detected face regions of the past frame to track the faces. The tracking method is superior in the stability of the regions but it is inferior in the face tracking efficiency. We proposed a face tracking method using the area of the overlapping region in the detected face regions of the past frame and the present frame to improve the tracking efficiency. The proposed face tracking efficiency demonstration was performed by making a film of face detection with face tracking in real-time and using the moving traces of the detected faces.