Hee-chul Han

Automatic illumination and color compensation using mean shift and sigma filter

We present a novel framework for automatic illumination and color compensation algorithm using mean shift and the sigma filter (ICCMS) to restore distorted images taken under the arbitrary lighting conditions. The proposed method is effective for appropriate illumination compensation, vivid color restoration, artifacts suppression, automatic parameter estimation, and low computational cost for HW implementation. We show the efficiency of the mean shift filter and sigma filter for illumination compensation with small sized kernel while considering the processing time and removing the artifacts such as HALO and noise amplification. The proposed color restoration function can restore the natural color and correct color noise artifact more perceptually compared with conventional methods. For the automatic processing, the image statistics analysis estimates suitable parameter and all constants are pre-defined. We also introduce the ROI-based parameter estimation dealing with small shadow area against spacious well-exposed background in an image for the touch-screen camera. The object evaluation is performed by CMC, CIEde2000, PSNR, SSIM, and 3D CIELAB gamut with state-of-the-art research and existing commercial solutions.

Virtual out of focus with single image to enhance 3D perception

The color, contrast and detail of an object in focus are accentuated, whereas those of a background are attenuated based on the estimated depth map with a face detection and segmentation method for enhancing a 3D perception. Considering the human perception and a real out of focused image taken with wide aperture lens, we suggest a boundary gradation for the estimated depth map to handle blurring errors. To make a blurred background, we perform a modified Gaussian pyramid by scaling up and blending all of images.

Quadratic formulation of disparity estimation problem for light-field camera

Newly available light-field (LF) cameras are able to capture several views of a scene simultaneously. These views typically have small parallax, and thus can be easily registered. In this paper we exploit this property of the views captured by the LF camera to formulate disparity estimation problem as a quadratic energy minimization problem. Our problem formulation has three benefits. Firstly, it allows computation of continuous disparity with subpixel accuracy. Secondly it permits recovering disparity of loosely textured objects and ensures that the disparity boundaries are aligned with the objects boundaries. And, finally, it allows finding the solution very quickly. It takes 15-20s for our non-optimized Matlab code to compute the solution for 25 × 350 × 350 input views.

Alternating line high dynamic range imaging

In this paper, a novel method of improving dynamic range of imaging device is proposed. Unlike multi-capture HDR methods, the suggested method requires only a small modification to existing CMOS imaging sensor. By changing exposure time line by line, one can obtain a multi-exposure image in a single capture of an image. With such captured image, digital image processing algorithms are applied to recover the lost vertical resolution with minimal degradation. This paper will cover overall system issues as well as detailed description of the algorithms applied to recover high dynamic range image.

Face relighting based on virtual irradiance sphere and reflection coefficient

We present a novel method to estimate the light source direction and relight a face texture image of a single 3-D model under arbitrary unknown illumination conditions. We create a virtual irradiance sphere to detect the light source direction from a given illuminated texture image using both normal vector mapping and weighted bilinear interpolation. We then induce a relighting equation with estimated ambient and diffuse intensity. We provide the result of a series of experiments on light source estimation, relighting, and face recognition to show the efficiency and accuracy of the proposed method in restoring the shading and shadow areas of a face texture image. Our approach for face relighting can be used for illuminant-invariant face recognition applications.

3D perception enhancement using depth map based color processing

The color, contrast and detail of an object in focus are increased, whereas those of a background are decreased based on the depth map for simulating atmospheric perspective and proximity luminance.

Illumination Estimation Based Color Contrast Enhancement

We present the framework for a color contrast enhancement using an illumination estimation, color balancing and color dynamic range expansion based on characteristics of a object reflectance, effect of illumination and human face color depending on human race. The method aims to emulate the way in which the human visual system discriminates original color and opposite color for increasing color contrast. The proposed method is effective for vivid color restoration and suppression of artifacts by proposed parameter weighting control under various illumination conditions.

Texture preserving noise reduction for ultra high ISO images

We present a novel method to reduce a sparkle and impulse noise of high ISO images. We employ diverse approach for a intensity(Y) image and two Color images (CB,CR) separately. For the intensity image, we propose a frequency cut based texture analysis to distinguish between noise patch and texture patch. A noise analysis calibration is proposed depending on exposure time, ISO and sensor size for the parameter extraction. The intensity noise is removed by a frequency magnitude cut based noise reduction assuming that the pattern patch has much higher magnitude than noise patch. However the color parts produce big size of noise mass after CFA interpolation. We proposed a coincidental smoothing filter for removing the color noise mass by applying the Gaussian function to the strength of the color. Finally we blend an input noisy image to a noise reduced image and add Gaussian noise for making a natural image and reducing some pattern loss.

Lens simulation with light field camera

We introduce the post processing method for virtual lens which simulates the focal length with super resolution and image stitching algorithm as well as the aperture with light-field rendering and lightness perception. Experimental results show the effectiveness of the cost less reconstruction technique for handling plenty of ordinary lenses.

HVS-aware ROI-based illumination and color restoration

We present the framework for Human visual system (HVS)-aware Region of interest (ROI)-based Illumination and Color Restoration called ‘HRICR’ for restoring distorted images taken under the arbitrary illumination environment. The proposed method is effective for appropriate illumination and vivid color restoration as well as suppression of artifacts such as HALO and noise amplification. We introduce the ROI-based parameter estimation dealing with small shadow area against spacious well-exposed background in an image for the touch-screen camera. The perceptual difference threshold is obtained using just-noticeable-difference (JND) based on HVS in response to user interaction.