SuGil Choi

A method for counting moving and stationary people by interest point classification

While the methods for people counting based on moving interest points have been shown good performance, the problem related to count static or temporarily stopped people remains particularly challenging. This paper presents a novel method for people counting which is considering moving and stationary people. The proposed method first separates the moving and static points by the motion information. Then, the temporarily static points of stationary people classified by analyzing the texture information of the points between the current and the eigenbackground image. Finally, we estimate the number of people using the moving and the temporarily static points. The experimental results show that the proposed method can identify the static points related to people correctly. Additionally, the results confirm that the proposed method reduces the estimation error in the sequences containing the many temporarily static people.

On the Challenges of Applying Selective Encryption on Region-of-Interest in H.264 Video Coding

Selective encryption of video data is not a new idea and various techniques proposed in previous literatures can effectively protect sensitive data from unauthorized access. After the initial appearance of selective encryption, some principles have become well known and widely adopted. One of the principles is the standard format compliance, which means that unauthorized users can decode and process the encrypted video without knowing the content. In the context of ROI encryption, this allows unauthorized users to decode and view the video except for the encrypted region containing sensitive information. However, there are some challenges for providing format compliance in H.264 video coding when applying encryption on certain regions and it is very important to address this issue in some applications such as privacy masking in video surveillance system. In this work, we describe the challenges and introduce some approaches to address these. Due to the bulky size, applying standard ciphers to multimedia data tends to be inefficient for real-time processing scenarios. Therefore, selective encryption initially appeared as means to relieve computational cost and some principles for designing selective encryption scheme have become widely adopted. Those principles include security, time efficiency, format compliance, compression performance and error robustness. The format compliance means that the encryption process does not change the encoded bitstreams data format in order to support such direct operations as browsing, playing, cutting, copying and so on. In the past decade, several selective encryption algorithms supporting format compliance have been reported, most of which are based on MPEG and H.264 video codec. In the context of ROI encryption, format compliant encryption algorithm allows unauthorized users to decode and view the video except for the encrypted region containing sensitive information. In the application of video surveillance system, the problem of privacy invasion can be addressed by encrypting only a privacy sensitive region and, if the encryption algorithm guarantees format compliance, a person with a lower level of security clearance can only get the non- sensitive information, and all the privacy information can be entirely unperceivable to him/her. Although ROI encryption supporting format compliance is simply the application of established selective encryption algorithm on a given region, ROI encryption in currently prevailing video codec, such as H.264 and MPEG4, has a number of unique challenges. The challenges are all about confining the degradation of visual quality to the specified region and keeping the other parts intact. However, most of the consisting elements in the encoded video do have little meaning in itself and can reconstruct a valid data only when combined with other elements. When a consisting element outside a ROI is dependent upon the element inside a ROI and the related element is encrypted, the reconstructed value outside a ROI can be unexpected one and can introduce artifacts in decoded video. As every video codec specifies different encoding algorithm, the challenges are dependent on the employed codec. In case of MPEG4, two challenges are identified and the solutions are described in (1). Coding schemes for independent ROI encryption in scalable video coding are presented in (2).

Implementation of a TCG-Based Trusted Computing in Mobile Device

Our implementation is aimed at estimating the possibility of employing TCG-based trusted computing mechanisms, such as verifying the code-integrity of executables and libraries at load-time and remote attestation, in mobile devices. Considering the restrained resource in mobile device, the experimentation shows promising results, thereby enabling these mechanisms to be used as a basic building block for a more secured mobile service. To this end, we add a new feature of integrity measurement and verification to Wombat Linux kernel and Iguana embedded OS. We also implement attestation agents, Privacy CA, and TCG Software Stack.

Improvement on TCG attestation and its implication for DRM

TCG (Trusted Computing Group) has defined a set of standards. The main features of the standards are protection against theft of secrets held on the platform and a mechanism for the platform to prove that it is in a trusted state, called attestation. However, the attestation mechanism is vulnerable to relay attack because of the lack of linkage between the endpoint identity and attestation message. We show here how to defeat the attack by employing a new agent, called Network Interface Monitoring Agent (NIMA). In addition, we show that the NIMA-based approach can render DRM more robust and efficient, especially in case of protecting a companys sensitive data.

New binary descriptors based on BRISK sampling pattern for image retrieval

Over the last decade, feature point descriptors such as SIFT have become indispensable tools in the computer vision community. But, the descriptors high computational overhead becomes a significant concern when it has to be on a device with limited computational and storage resources. In order to make descriptors faster to compute and more compact, several binary descriptors such as ORB and BRISK have been proposed. These binary descriptors are not successful in image retrieval, so we propose new binary descriptors to increase the accuracy while maintaining computational efficiency.

Evaluating weighting schemes for adult image detection using bag of visual words

As a growing number of minors are exposed to pornographic multimedia, the need to prevent inappropriate access to this has increased. Adult image detection i s a n important technology for this purpose, and several algorithms have been proposed. However, the simple methods such as analyzing skin regions or global histogram suffer from low classification accuracy. Recent advances in adult image detection used bag of visual words approach and showed promising result. In this paper, we experiment with several weighting schemes for visual words histogram construction and show that classification accuracy can be increased.

Fine-level moving object detection using region-based background/foreground model

Background subtraction forms the first stage in an automated visual surveillance system. Most fine-level background subtraction techniques are based on the analysis of pixel-based distribution, while region-based models are more robust to dynamic backgrounds. Hence, it is reasonable to refine the coarse foreground extracted by region-based subtraction, so that detailed shape information can be available just like the result of pixel-based subtraction. In this paper, we propose a method for the fine processing of coarse foreground block.

A comparison of keypoint detectors in the context of pedestrian counting

The existing pedestrian counting methods use the various keypoint detectors but there is no attempt to find a suitable keypoint detector for counting pedestrians. Therefore, in this paper, we compare the various keypoint detectors using a public dataset. Our evaluation framework uses the processing time of keypoint detection and matching as a performance measure. Also, we use the accuracy of moving points which means the percentage of the number of moving points associated persons in all moving points. Experimental results show that the BRISK method is more adequate for pedestrian counting method because of its low processing time and high accuracy.

Fast image retrieval with grid-based keypoint detector and binary descriptor.

As an alternative to vector-based descriptors, such as SIFT and SURF, more computationally efficient binary descriptors, such as BRISK and ORB, have recently been proposed. These binary descriptors are usually used in combination with a novel scale-space FAST-based detector to be suitable for real-time applications, but it consumes more time than creating binary descriptors. Therefore, if accuracy can be kept similar, keypoint sampling by a grid is better than FAST-based detector because it consumes almost no time. In this paper, grid-based sampling and BRISK keypoint detector are tested for image retrieval. Experimental results demonstrate that grid-based sampling out performs keypoint detector in terms of accuracy and processing speed.

Privacy Preservation in SAT (Single Authentication Through)

In SAT (Single Authentication Through) scheme, each smart camera is capable of identifying, tracking identified objects, and delivering associate ID information to sibling subjects. So in this paper, we suggest a privacy preservation scheme for preventing privacy infringement during propagation of sensitive information.