Si-Woong Lee

Spatio-temporal video segmentation using a joint similarity measure

This paper presents a new morphological spatio-temporal segmentation algorithm. The algorithm incorporates luminance and motion information simultaneously and uses morphological tools such as morphological filters and watershed algorithm. The procedure toward complete segmentation consists of three steps: joint marker extraction, boundary decision, and motion-based region fusion. First, the joint marker extraction identifies the presence of homogeneous regions in both motion and luminance, where a simple joint marker extraction technique is proposed. Second, the spatio-temporal boundaries are decided by the watershed algorithm. For this purpose, a new joint similarity measure is proposed. Finally, an elimination of redundant regions is done using motion-based region fusion. By incorporating spatial and temporal information simultaneously, we can obtain visually meaningful segmentation results. Simulation results demonstrates the efficiency of the proposed method.

Rate control of MPEG video for consistent picture quality

We propose a rate control scheme using a rate-distortion (R-D) estimation model, which produces a consistent picture quality between consecutive frames. Our R-D estimation method offers a closed-form mathematical model that enables us to predict the bits and the distortion generated from a frame encoded at a given quantization parameter (QP) and vice versa. Its most attractive feature is its low computational complexity. Furthermore, it is accurate enough to be applied to practical video coding. In our simulation, the estimation errors for rate and distortion are less than 2.5% and 1.5%, respectively. Therefore, the proposed rate control scheme is appropriate for applications requiring low delay, low complexity, and the ability to control output bit-rate and quality accurately. Our scheme ensures that the video buffers do not underflow or overflow by satisfying the buffer constraint, and it also prevents quality difference between consecutive frames from exceeding a certain demanded level by adopting a distortion constraint. In addition, a consistent picture quality is maintained within a frame, and error propagation, caused by quality degradation of anchor frames, is reduced by differentiating the control procedure for anchor frames from that for nonanchor frames. Simulation results show that our control scheme achieves 0.52-1.84 dB peak signal-to-noise ratio (PSNR) gain over MPEG-2 Test Model 5 (TM5) rate control and maintains very consistent quality within a frame as well as between frames.

Fast motion estimation with adaptive search range adjustment

A fast motion estimation algorithm based on the adaptive search range adjustment (ASRA) is presented. The size of search range for each block is adaptively controlled according to the initial cost at the search center, and then the full search algorithm (FSA) is performed to find the optimal matching block in the adjusted search range. The proposed method is implemented in H.264/AVC reference software (JM) to demonstrate its efficiency over the conventional FSA.

Detection and classification of hypovolaemia during anaesthesia

In recent years, there has been a rapid growth in patient monitoring and medical data analysis using decision support systems, smart alarm monitoring, expert systems and many other computer aided protocols. The main goal of this study was to enhance the developed diagnostic alarm system for detecting critical events during anaesthesia. The proposed diagnostic alarm system is called Fuzzy logic monitoring system-2 (FLMS-2). The performance of the system was validated through a series of off-line tests. When detecting hypovolaemia a substantial level of agreement was observed between FLMS-2 and the human expert and it is shown that system has a better performance with sensitivity of 94%, specificity of 90% and predictability of 72%.

Video rate control using an adaptive quantization based on a combined activity measure

A new rate control algorithm for videos is presented. The method comes from the MPEG-2 Test Model 5(TM5) rate control, while a buffer constraint and a new measure for the macroblock (MB) activity based on spatio-temporal sensitivity are introduced. Experimental results show that the proposed method outperforms the TM5 rate control in picture quality.

Error resilient texture coding scheme for wireless Video transmission based on coefficient sampling and interleaving

We tackle the problem of burst error in wireless video trans- mission. Burst error damages video information so severely that it is very difficult to restore the lost information. For restoration, it is advantageous to retain a part of block information. To achieve this, we propose a coef- ficient sampling and interleaving method, where texture coefficients are grouped into several coefficient groups CGs so that the groups have equal importance. The CGs are interleaved so that burst error cannot damage all CGs from one block. A restoration method is proposed to utilize the retained coefficients. The simulation results show that the pro- posed coefficient sampling scheme and the restoration method restore the lost information very well.

Robust change detection by global-illumination-change compensation and noise-adaptive thresholding

Robustness to noise and/or illumination variations has been a major issue with change detection. We present a novel change detection scheme that is robust to both noise and illumination changes. The major difference of our method compared to the existing ones is that our algo- rithm globally estimates and compensates for the illumination changes between two input images, whereas the previous approaches rely on some local change metrics for illumination-independent detection. Spe- cifically, a piecewise quadratic model for describing global illumination changes (GICs) is formulated, and an efficient estimation algorithm for the model parameters is proposed. In this way, we can successfully de- tect any types of local changes, whereas the previous approaches show either limited performances in detecting object interiors or severe perfor- mance degradations when global illumination changes occur. We also present a noise-adaptive thresholding method for the GIC-compensated intensity differences. The noise variance is estimated accurately based on symmetricity and decrease of zero-mean noise distributions, and then the decision threshold is selected adaptively to the estimated variance. Experimental results show that the proposed method is very robust to both noise and illumination changes, and outperforms the previous algo- rithms in various aspects.

Frame rate up-conversion method based on texture adaptive bilateral motion estimation

In this paper, a new frame rate up-conversion scheme is proposed in order to overcome motion blur problems of liquid crystal displays caused by slow responses. Conventional bilateral motion estimation methods used in the frame rate upconversion scheme present a major drawback of missing true motion vectors if there are blocks with simple textures in search range. To solve this problem, a texture adaptive bilateral motion estimation method increasing the cost value of a block with simple texture is proposed. Also a motion estimation scheme that utilizes neighboring motion vector effectively is proposed to reduce computation time required to estimate motion. Since the proposed scheme does not apply all available motion vectors within search range, the execution time of frame rate up conversion can be reduced dramatically. Experimental results show that the interpolated frame image quality is improved as well as the processing time of the proposed method is greatly reduced compared with those of conventional methods.

Adaptive Depth Fusion based on Reliability of Depth Cues for 2D-to-3D Video Conversion

3D video is regarded as the next generation contents in numerous applications. The 2D-to-3D video conversion technologies are strongly required to resolve a lack of 3D videos during the period of transition to the full ripe 3D video era. In 2D-to-3D conversion methods, after the depth image of each scene in 2D video is estimated, stereoscopic video is synthesized using DIBR (Depth Image Based Rendering) technologies. This paper proposes a novel depth fusion algorithm that integrates multiple depth cues contained in 2D video to generate stereoscopic video. For the proper depth fusion, it is checked whether some cues are reliable or not in current scene. Based on the result of the reliability tests, current scene is classified into one of 4 scene types and scene-adaptive depth fusion is applied to combine those reliable depth cues to generate the final depth information. Simulation results show that each depth cue is reasonably utilized according to scene types and final depth is generated by cues which can effectively represent the current scene.

Reliable watermark detection method based on analysis of correlation

A digital watermark is a perceptually unobtrusive signal em- bedded in a multimedia asset, such as an image, for copyright protec- tion. In many cases watermark detection amounts to thresholding a cor- relation value between a watermark and a received image. Watermarking detection schemes can be classified into two types. Type 1 is based on a correlation process that is applied to the difference between an original image and an input image to be tested. Type 2 is based on a correlation process that is directly applied to an input image. Type 1 fails to prove the rightful ownership, while type 2 can do so. However, type 2 has a problem that type 1 does not have: The correla- tion between a watermark and an original image to be watermarked is too significant to be ignored, when it is normalized by the watermarks energy. In this paper, based on an analysis of the correlation, we pro- pose a novel watermarking scheme to minimize that effect. We verify the performance of the proposed scheme by experiments.