Wonjun Kim

Background Subtraction Using Illumination-Invariant Structural Complexity

In this letter, we propose a novel method for background subtraction in outdoor scenes. Inspired by the observation that the orthogonal decomposition onto a set of pixel intensities efficiently reveals illumination effects, we exploit a simple, yet powerful feature for describing the underlying structure of the local region in a given video, the so-called illumination-invariant structural complexity (IISC). In contrast to previous approaches still suffering from high-level false positives driven by varying illuminations in outdoor environments, our IISC feature has an ability to greatly discriminate structural changes by moving objects from those by illumination effects. We also provide the theoretical analysis to confirm that the proposed IISC feature is useful for modeling the background under diverse lighting conditions. Moreover, our framework does not require any preprocessing task. Experimental results on various datasets demonstrate that the proposed method is effective for video surveillance in a wide range of outdoor environments.

Fingerprint Liveness Detection Using Local Coherence Patterns

In this letter, we propose a novel image descriptor for fingerprint liveness detection using the local coherence of a given image. Based on the observation that materials employed for making fake fingerprints (e.g., silicone, wood glue, etc.) tend to yield the nonuniformity in the captured image due to the replica fabrication process, we focus on the difference of the dispersion in the image gradient field between live and fake fingerprints. More specifically, we propose to define the local patterns of the coherence along the dominant direction, the so-called local coherence patterns, as our features, which are fed into the linear support vector machine (SVM) classifier to determine whether a given fingerprint is fake or not. Experimental results on various datasets show that the proposed image descriptor is effective for fingerprint liveness detection compared to other approaches employed in the literature.

Illumination-Invariant Background Subtraction: Comparative Review, Models, and Prospects

Background subtraction is a key prerequisite for a wide range of image processing applications due to its pervasiveness in various contexts. In particular, video surveillance highly requires the reliable background subtraction for further operations, such as object tracking and recognition, and thus, enormous efforts for this task have been devoted in recent decades. However, the path of technological evolution for background subtraction has now faced with an important issue that has started to be resolved: sensitivity to dynamic changes of scene contexts (e.g., illumination variations and moving backgrounds). Such dynamic changes are hardly tolerated by most of traditional background models, since they yield the drastically different statistics of pixel values even onto the relevant position between consecutive frames. To resolve this problem, many researchers in this field have developed robust and efficient methods. The goal of this paper is to provide a comprehensive review with a special attention to schemes related to handling varying illuminations frequently occurring in the outdoor surveillance scenario. This paper covers a systematic taxonomy, methodologies, and performance evaluations on benchmark databases, and also provides constructive discussions for the smart video surveillance under unconstrained outdoor environments.

Directional coherence-based spatiotemporal descriptor for object detection in static and dynamic scenes

This paper presents a simple, yet powerful local descriptor, so-called the histograms of space–time dominant orientations (HiSTDO). Specifically, our HiSTDO is composed of two main components, i.e., the dominant orientation and its coherence, which represents how intensively gradients in the local region are distributed along the space–time dominant orientation. By incorporating them into the histogram, we define it as our HiSTDO descriptor. In contrast to previous methods vulnerable to the presence of the background clutter and the camera noise, our HiSTDO greatly encodes the space–time shape of underlying structures even under such challenging conditions, and it can thus be efficiently applied to various applications (e.g., object and action detection). Experimental results on diverse datasets demonstrate that the proposed descriptor is effective for human action as well as object detection.

Background subtraction with variable illumination in outdoor scenes

Background subtraction is a key prerequisite for intelligent video surveillance, but most of the methods employed are still affected by dynamic changes in the illumination conditions, e.g., shadows cast by passing clouds occur frequently in outdoor scenes. To resolve this problem, a novel approach based on the underlying structure of the difference image is introduced in this study. In particular, local binary patterns (LBPs) are computed based on the frame differencing result, i.e., moving LBP, and then compared with the background model, which is updated according to an online interpolation scheme, in order to determine whether the current pixel belongs to the background. An important advantage of the proposed method is that it efficiently smoothes unexpected noise between frames while also preserving the boundaries of the moving objects by using an edge-aware filtering technique. Experimental results obtained using two benchmark data sets demonstrated that the proposed method is more robust to variable illumination in outdoor scenes compared with previously proposed approaches.

Multiple player tracking in soccer videos: an adaptive multiscale sampling approach

Visual tracking is an essential technique in computer vision. Even though the notable improvement has been achieved during last few years, tracking multiple objects still remains as a challenging task. In this paper, a novel method for tracking multiple players in soccer videos, which include severe occlusions between players and nonlinear motions by their complex interactions, is introduced. Specifically, we first extract moving objects (i.e., players) by refining results of background subtraction via the edge information obtained from the frame differencing result. Then, we conduct multiscale sampling in foreground regions, which are spatially close to each tracked player, and subsequently computing the dissimilarity between sampled image blocks and each tracked player. Based on the best-matched case, the state of each tracked player (e.g., center position, color, etc.) is consistently updated using the online interpolation scheme. Experimental results in various soccer videos show the efficiency and robustness of our method compared to previous approaches introduced in the literature.

Moving object detection using edges of residuals under varying illuminations

This paper presents a new method for moving object detection under varying illuminations. The key idea of the proposed method is to reconstruct moving objects from edges computed on the result of frame differencing, the so-called edges of residuals. This scheme forces lighting variations to be efficiently suppressed in the gradient space while preserving the boundary of moving objects. The inner areas of moving objects are subsequently reconstructed by utilizing image gradients of the original frame, which are masked by edges of residuals, in a least-square sense. One important advantage of the proposed method is to uniformly highlight moving objects regardless of their scales. Experimental results on various databases demonstrate that the proposed method is effective for moving object detection under diverse lighting conditions.

A comparative study on multi-object tracking methods for sports events

Due to the rapid growth of machine learning technology, there is a need for research to automatically recognize objects and analyze their behavior in various fields, as is the case with sports. Currently, a system for detecting and tracking multiple objects in a sporting event is not accurate enough. Since most of the services depend on the manual operation of an experienced operator, it is necessary to develop a real time tracking technique for detecting the position of an object. In this paper, we propose an algorithm for multi-object tracking in a sporting event by presenting the results of comparing the performance of existing algorithms for multi-object tracking.

Towards real biometrics : An overview of fingerprint liveness detection

This paper presents an overview of liveness detection particularly based on fingerprint images. Since fingerprints start to be widely employed in the mobile devices such as smartphones and tablets for payment as well as security, high-performed algorithms have been explored in literature. At the same time, such mobile systems are highly required to detect spoofing attacks by fabricated fingerprints with malicious intends. To this end, many researchers have developed liveness detection methods based on local image descriptors. In this paper, we briefly review those descriptors used for fingerprint liveness detection and compare the detection performance based on representative datasets, i.e., ATVS, LivDet2009, and LivDet2011. From experimental results, we can conclude that textural patterns of a local region in the fingerprint image have a good ability to discriminate fake fingerprints from live ones.