Mun Cheon Kang

Improved simple linear iterative clustering superpixels

The superpixels are small regions in an image which do not contain edges inside. Among all the superpixel algorithms, the simple linear iterative clustering (SLIC) method is widely adopted due to its practicality. However, the resultant superpixels sometimes do not well adhere to the edges. In this paper, we present an improved SLIC superpixel method to solve this problem. Experiments show that the proposed method produces superpixels whose boundaries better adhere to the image edges without significantly increasing the processing time compared with the conventional method.

A robust obstacle detection method for robotic vacuum cleaners

Conventional robotic vacuum cleaners (RVCs) with ultrasonic or infrared (IR) sensors present problems in detecting obstacles when they clean the floor in complex situations, for example, under tables or chairs with thin legs. This paper presents a robust obstacle detection (OD) method based on the triangulation principle for RVCs operating in various home environments. The proposed method uses the IR emitter of the RVC to project a horizontal IR beam toward the floor, following which the RVCs wide-angle vision camera captures an image that includes the IR line reflected by the floor or an obstacle. Obstacles are detected by using the image coordinates of the pixels that belong to the IR line in the captured image. Accurate separation of the IR line from the image background is accomplished by defining and minimizing an energy function based on the characteristics of the IR line. The proposed method was tested on the embedded RVC system and was shown capable of achieving OD performance compared with existing methods.

An edge-guided image interpolation method using Taylor series approximation

In this paper, a novel real-time edge-guided interpolation method is presented to produce the high resolution image without any disturbing artifacts such as blurring, jagging, and overshooting. The proposed method computes the first- and second-order derivatives of an input image to measure the geometry of edges in the image. Based on these measures, the value of a pixel to be interpolated is estimated along four directions using Taylor series approximation. The four directional estimates are adaptively fused based on the orientation of a local edge to obtain the edge-guided interpolation output. Experimental results demonstrate that the proposed interpolation method outperforms the conventional ones both in terms of objective and subjective image qualities with low computational complexity.

A novel obstacle detection method based on deformable grid for the visually impaired

To assist visually impaired people, a variety of obstacle detection (OD) methods with the monocular vision have been developed. Most conventional OD methods detect the obstacle by using the variation of the motion vector or the object size between two consecutive image frames. However, the OD performance using the short-term variation is significantly affected by the tracking error, which leads to inaccurate OD results. To cope with this problem, this paper presents a novel OD method based on a new structure, called Deformable Grid (DG). The DG is initially a regular gridshaped, but it can be deformed gradually depending on the motion of the object in the scene. The proposed method detects the object at risk of collision based on the degree of the deformation of the DG. Experimental results demonstrate that the proposed OD method outperforms the conventional ones in terms of processing time as well as accuracy.

A novel depth image enhancement method based on the linear surface model

In three-dimensional (3D) video applications, structured-light RGB-D cameras are commonly used to capture depth images that convey the per-pixel depth information in a scene. However, these cameras often produce regions with missing pixels (MPs). These regions, referred to as holes, will not contain no any depth information for the captured depth image. In this paper, a novel depth image enhancement method that accurately estimates depth values of MPs is presented. In the proposed method, the neighboring region outside the hole is first segmented into superpixels using simple linear iterative clustering. Subsequently, the depth value trend of each superpixel is modeled as a linear surface. Finally, one of the linear surfaces is selected using a proposed metric, to estimate the depth value of a particular MP in the hole. Experimental results demonstrate that the proposed method provides superior performance, especially around the object boundary, compared with other state-of-theart depth image enhancement methods.

An enhanced obstacle avoidance method for the visually impaired using deformable grid

Various obstacle avoidance (OA) methods with the monocular vision have been studied to assist visually impaired people. Recently, a method that effectively locates obstacles at a risk of collision using the shape variation of a grid, called deformable grid (DG), was introduced [11]. To further improve the detection performance of the previous work, a vertex deformation function representing the displacement of each vertex in the DG is firstly defined employing the perspective projection geometry. Then, the collision risk is accurately estimated based on the shape variation of the DG that is measured using the vertex deformation function. Experimental results show that the proposed method outperforms other conventional methods as well as the previous work in terms of both the accuracy and processing time, and clearly detects obstacles that present a risk of collision in a variety of scenarios. This method is suitable for application to electronic travel aid (ETA) systems using consumer devices, such as smart-phones, tablets, and wearable devices.

Collision detection based on scale change of image segments for the visually impaired

A variety of portable or wearable navigation systems mounted on smart glasses and smartphones have been developed to assist visually impaired people over the last decade. In these systems, collision detection is one of the key components. Many conventional methods with the monocular vision estimate the collision risk based on the motion information of obstacles in the image by measuring the size change of objects using detected feature points and their corresponding motion vectors. However, the size change is sometimes incorrectly measured due to unreliable feature points and motion vectors. To overcome this problem, we present a motion clustering scheme to remove outliers among both feature points and motion vectors. Experimental results indicate that the proposed collision detection method outperforms the conventional one in terms of detection and false positive rates.

An improved GrabCut using a saliency map

The GrabCut, which uses the graph-cut iteratively, is popularly used as an interactive image segmentation method since it can produce the globally optimal result. However, since the initialization of the GrabCut is roughly performed by the manual interaction, the accuracy of the segmentation result is not guaranteed when the user defines an inaccurate guide. To solve this problem, in this paper, we present an improved GrabCut method which uses a visual saliency of the target region for the effective initialization. Experimental results demonstrate that the proposed method provides more accurate segmentation results compared with the conventional method.

High dynamic range image tone mapping based on asymmetric model of retinal adaptation

Abstract Global tone mapping operators using the symmetrical retinal response model to light tend to produce a low dynamic range (LDR) image that exhibits loss of details of its corresponding high dynamic range (HDR) image in a bright or dark area. In this paper, we introduce a new asymmetric sigmoid curve (ASC) based on the model of retinal adaptation encompassing symmetrical S-shaped curve, and present two global tone mapping operators by using the ASC. In the proposed method, an ASC-based tone mapping function is obtained by using a well-known classic photography technique, called zone system. In addition, a contrast-enhancing tone mapping function is introduced by formulating a bi-criteria optimization problem with the luminance histogram of an input HDR image and the ASC-based mapping function. Experimental results demonstrate that the proposed method enhances the global contrast while preserving image details in the tone-mapped LDR image. Moreover, the objective assessment results using an image quality metric indicate that the proposed method shows a high performance to state-of-the-art global tone mapping operators.

A novel burn-in potential region detection method using image processing technique

The organic light emitting diode (OLED) display has been widely adopted to various multimedia devices with superior performance in terms of image quality and power efficiency. However, the luminance degradation of the OLEDs, called burn-in, is still one of the major problems. This paper presents a novel method of detecting the burn-in potential region (BPR) to alleviate the luminance degradation. First, in order to extract the burn-in potential pixels (BPPs) which deteriorate the uniformity of the display, we calculate the remaining lifetime of OLED of each pixel. Then, the BPRs are detected by the level set based image segmentation using the BPPs as the seed points. The experimental results demonstrate that the proposed method detects BPRs with superior effectiveness compared with other alternative methods.