Bongjoe Kim

Mahalanobis Distance Cross-Correlation for Illumination-Invariant Stereo Matching

A robust similarity measure called the Mahalanobis distance cross-correlation (MDCC) is proposed for illumination-invariant stereo matching, which uses a local color distribution within support windows. It is shown that the Mahalanobis distance between the color itself and the average color is preserved under affine transformation. The MDCC converts pixels within each support window into the Mahalanobis distance transform (MDT) space. The similarity between MDT pairs is then computed using the cross-correlation with an asymmetric weight function based on the Mahalanobis distance. The MDCC considers correlation on cross-color channels, thus providing robustness to affine illumination variation. Experimental results show that the MDCC outperforms state-of-the-art similarity measures in terms of stereo matching for image pairs taken under different illumination conditions.

Exact order based feature descriptor for illumination robust image matching

We present a novel method for a feature descriptor called an exact order based descriptor (EOD). The proposed method consists of three steps. First, to resolve ordering ambiguity for pixels of the same intensity, an exact order image is created by changing the discrete intensity into a k-dimensional continuous value. Second, exact order based features are generated globally and locally. Finally, the EOD is constructed by combining the global and local exact order features using the discrete cosine transform. Experimental results show that the proposed method outperforms other state-of-the-art descriptors over a number of images.

Illumination invariant road detection based on learning method

Road detection is an essential and important component in intelligent transportation system (ITS). Generally, most road detection methods are sensitive to variation of illumination which results in increasing false detection rate. In this paper, we propose an illumination invariant road detection method to deal with variation of illumination. We adopt learning method to estimate illumination invariant direction which is specified to road surface. Once this direction is estimated, we can classify image pixel as road or not. Incorporating scene layout of road image, we reduce false positive detection rate outside the road. Experimental results on real road scenes show that the effectiveness of the proposed method.

Illumination-invariant color space and its application to skin-color detection

Color-based digital image processing (DIP) techniques have attracted much attention in many vision-based applications. However, due to color variations resulting from illumination changes, many color-based DIP techniques have yet to demonstrate a stable state of performance. Skin-color detection, which is one of the popular color-based DIP techniques, must overcome the illumination problems. We address the issue by presenting an illumination-invariant color space based on the image acquisition model that is determined by the Lambertian surface. Furthermore, we propose a method of skin-color detection based on the illumination-invariant color space. To evaluate the performance in terms of the illumination-invariant property, we perform a skin-color detection experiment. In the experiment, we compare the proposed method with the methods based on several color spaces. From the experiment, we achieve encouraging results, and our empirical experiments evidence both the effectiveness and the usefulness of the proposed method.

CIECAM02-based tone mapping technique for color image contrast enhancement

We propose a CIECAM02-based tone mapping technique for color image contrast enhancement that works especially well in bright surrounding conditions. The CIECAM02 is inherently strong when considering human visual systems (HVS) and surrounding conditions. In general, dark regions look darker in bright surrounding conditions, so some details in these dark regions tend to disappear. To overcome this problem, we propose a global tone mapping technique for color image contrast enhancement that uses a gamma encoding and the I/O relationship of the CIECAM02 in bright surrounding conditions. The proposed method makes the contrast of the source image stronger in bright surrounding conditions. We confirmed the superior performance of the proposed method in bright surrounding conditions by psychophysical experiment.

Illumination invariant skin color segmentation

Skin color segmentation takes a great attention in many vision-based methodologies. However, the performance of the segmentation is not stable because of color variations caused by various illumination conditions. In this paper, we propose a new skin color model for segmentation which is invariant to illumination variation. It is based on illumination-free color space which is defined on Lambertian surface. Through several experimental results, we confirm that the proposed skin color model is successfully applied to skin color segmentation under various illumination conditions.

Robust Corner Detection Based on Image Structure

In this paper, we propose a robust corner detection method to improve both detection rate and localization accuracy by modifying the structure tensor-based corner detection method in two ways. First, we introduce a connected component analysis (CCA) method for constructing a CCA structure tensor in order to make the structure tensor adaptive to the structure of the image. Second, the normalized cross-correlation (NCC) method is applied for false corner rejection with the observation that the patch of a true corner has a distinctive characteristic compared with connected neighboring patches. The proposed method is compared with several corner detection methods over a number of images. Experimental results show that the proposed method shows better performance in terms of both detection rate and localization accuracy.

Surrounding adaptive color image enhancement based on CIECAM02

In this paper, we propose a CIECAM02-based color image enhancement method which is particularly robust to scenes with bright surrounding. The proposed method detects color edges using a distance metric based on the characteristics of a human visual system (HVS). The CIECAM02 is inherently strong considering both HVS and surrounding conditions. A major problem of HVS is that the dark region appears darker under a bright surrounding condition, leading to masking of details within the dark region. This phenomenon causes deterioration of edges which are among the most important and sensitive components for the HVS. To overcome this problem, we propose to weight the deteriorated edges at bright scenes. Adaptively, we estimate the surrounding image by the CIECAM02, and then use a vector gradient edge detector with a newly proposed distance metric to perform the weighting. The proposed method is seen to enhance edges without introducing unwanted color artifacts. We subjectively confirm the performance with clearly enhanced images.

Automatic illumination invariant road detection with stereo vision

Road detection is an important task in intelligent transportation system (ITS). Variation of illumination is a major issue in most of the vision approaches, which causes the problem of false detection. In this paper, we propose an automatic illumination invariant road detection method with stereo vision. Disparity and illumination invariant road direction are estimated and the road probability map is calculated to decide whether each pixel belongs to road or not. Then, we utilize joint bilateral filter to refine the detected road region. Experimental results on real road scenes show that the proposed method outperforms the conventional methods.

Errata: Haar-like compact local binary pattern for illumination-robust feature matching

A novel method for local image descriptor called a Haar-like compact local binary pattern (HC-LBP) is presented, which is robust to Gaussian noise and illumination changes and reduces the dimension of features by modifying the local binary pattern (LBP) method in two ways. First, the binary Haar-like feature method is applied to calculate region-based rank order. The binary Haar-like feature method maintains only the ordinal relationship and not the difference of intensity. Using a region-based binary operation, the proposed local image descriptor becomes more robust to Gaussian noise and illumination changes. Second, a HC-LBP feature is generated by applying the binary Haar-like feature according to the edge class in order to reduce feature dimensions and increase the discriminating power of each feature. In the feature matching experiment, the proposed method outperforms the popular scale invariant feature transform, center-symmetric local binary pattern and center-symmetric local ternary pattern in the presence of illumination changes, Gaussian noise, image blurring, and viewpoint change. Compared with popular local feature descriptors, the proposed method could get more than 0.1 increases in recall. Also, the HC-LBP descriptor could be used for many computer vision applications such as image retrieval, face recognition, and texture recognition.