• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.10 no.5
• /
• pp.13-20
• /
• 2010

We proposed a new technique to perform fast video stabilization using integral image in this article. In the proposed technique, it evaluate local and global motion by the block matching using the generated integral image for each frame and compensate the motion like jitter. We made the various experimental jitter patterns to evaluate the effectiveness of the proposed technique and evaluated stabilization capability and execution time with the existing ones. Through the experiment, we found that the execution time of proposed technique was faster than that of existing techniques and the compensation of jitter was well done.

• International journal of advanced smart convergence
• /
• v.5 no.4
• /
• pp.1-9
• /
• 2016

In order to reconstruct a 3D model in video sequences, to select key frames that are easy to estimate a geometric model is essential. This paper proposes a method to easily extract informative frames from a handheld video. The method combines selection criteria based on appropriate-baseline determination between frames, frame jumping for fast searching in the video, geometric robust information criterion (GRIC) scores for the frame-to-frame homography and fundamental matrix, and blurry-frame removal. Through experiments with videos taken in indoor space, the proposed method shows creating a more robust 3D point cloud than existing methods, even in the presence of motion blur and degenerate motions.

• Proceedings of the IEEK Conference
• /
• 2006.06a
• /
• pp.411-412
• /
• 2006

In this paper, we proposed a fast algorithm for inter prediction included the most complexity in H.264/AVC. It decide search range according to direction of predicted motion vector, and then perform adaptive candidate spiral search. Simultaneously, it perform motion estimation of variable loop with threshold for variable block size. Conclusively, it is implemented in JM FME with high complexity applying to rate-distortion optimization. Experimental results show that significant complexity reduction is achieved while the degradation in video quality is negligible.

• Journal of Sensor Science and Technology
• /
• v.7 no.6
• /
• pp.394-400
• /
• 1998

In this paper, we present a robust and fast method that enables real-time computing of axial motion component of different points of a scene from a stereo images sequence. The aim of our method is to establish axial motion maps by computing a range of disparity maps. We propose a solution in two steps. In the first step we estimate motion with a low level computing for an image point by a detection estimation-structure. In the second step, we use the neighbourhood information of the image point with morphology operation. The motion maps are established with a constant computation time without spatio-temporal matching.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.46 no.7
• /
• pp.53-60
• /
• 2009

Motion estimation for H.264/AVC video CODEC is very complex and requires a huge amount of computational efforts because it uses multiple reference frames and variable block sizes. We propose the architecture of high-performance integer-pixel motion estimation circuit based on fast algorithms for multiple reference frame selection, block matching, block mode decision and motion vector estimation. We also propose the architecture of high-performance interpolation circuit for sub-pixel motion estimation. We described the RTL circuit in Verilog HDL and synthesized the gate-level circuit using 130nm standard cell library. The integer-pixel motion estimation circuit consists of 77,600 logic gates and four $32\times8\times32$-bit dual-port SRAM's. It has tile maximum operating frequency of 161MHz and can process up to 51 D1 (720$\times$480) color in go frames per second. The fractional motion estimation circuit consists of 22,478 logic gates. It has the maximum operating frequency of 200MHz and can process up to 69 1080HD (1,920$\times$1,088) color image frames per second.

• The KIPS Transactions:PartB
• /
• v.11B no.3
• /
• pp.309-316
• /
• 2004

In this paper, we proposed the TDS(T-shape Diamond Search) based on the directions of above, below, left and right points to estimate the motion vector fast and more correctly in this method, we exploit the facts that most motion vectors are enclosed in a circular region with a radius of 2 fixels around search center(0,0). At first, the 4 points in the above, below, left and right around the search center is calculated to decide the point of the MBD(Minimum Block Distortion). And then w. above point of the MBD is checked to calculate the SAD. If the SAD of the above point is less than the previous MBD, this process is repeated. Otherwise, the right and left points of MBD are calculated to decide The points that have the MBD between right point and left point. Above processes are repeated to the predicted direction for motion estimation. Especially, if the motions of image are concentrated in the crossing directions, the points of other directions are omitted. As a result, we can estimate motion vectors fast. Experiments show that the speedup improvement of the proposed algorithm over Diamond Search algorithm(DS) and HEXgon Based Search(HEXBS) can be up to 38∼50％ while maintaining similar image Quality.

• Smart Media Journal
• /
• v.6 no.3
• /
• pp.49-56
• /
• 2017

Images and Videos that include the human face contain a lot of information. Therefore, accurately extracting human face is a very important issue in the field of computer vision. However, in real life, human faces have various shapes and textures. To adapt to these variations, A model-based approach is one of the best ways in which unknown data can be represented by the model in which it is built. However, the model-based approach has its weaknesses when the motion between two frames is big, it can be either a sudden change of pose or moving with fast speed. In this paper, we propose an enhanced human face-tracking model. This approach included human face detection and motion estimation using Cascaded Convolutional Neural Networks, and continuous human face tracking and modeling correction steps using the Active Appearance Model. A proposed system detects human face in the first input frame and initializes the models. On later frames, Cascaded CNN face detection is used to estimate the target motion such as location or pose before applying the old model and fit new target.

• Journal of Korea Society of Industrial Information Systems
• /
• v.13 no.1
• /
• pp.57-67
• /
• 2008

In this paper, we present an adaptive block matching motion estimation using optical flow. In the proposed algorithm, we calculate the temporal and spatial gradient value for each pixel value from tile differential filter, and estimate the optical flow which is used to decide the location and the size of the search region from the gradient values by least square optical flow algorithm. In particular, the proposed algorithm showed a excellent performance with fast and complex motion sequences. From the computer simulation for various motion characteristic sequences. The proposed algorithm shows a significant enhancement of PSNR over previous blocking matching algorithms.

• Journal of Korea Multimedia Society
• /
• v.18 no.4
• /
• pp.437-442
• /
• 2015

Voice activity detection is very important process that voice activity separated form noisy speech signal for speech enhance. Over the past few years, many studies have been made on voice activity detection, but it has poor performance in low signal to noise ratio environment or fickle noise such as car noise. In this paper, it proposed new voice activity detection algorithm using ensemble variance based on wavelet band entropy and soft thresholding method. We conduct a survey in a lot of signal to noise ratio environment of car noise to evaluate performance of the proposed algorithm and confirmed performance of the proposed algorithm.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2010.11a
• /
• pp.229-232
• /
• 2010

비디오 압축 기법에서 움직임 추정(Motion Estimation)은 매우 중요한 부분을 차지하는데, 그것은 움직임 추정이 화질과 인코딩 시간에 직접적으로 영향을 미치기 때문이다. 가장 기본적인 움직임 추정 기법은 전역 탐색 기법(Full Search)인데, 이는 가장 좋은 화질을 보여주긴 하지만 매우 많은 계산량이 필요하다는 단점이 있다. 따라서 좋은 화질을 유지하면서도 계산량을 낮추기 위한 많은 고속 탐색 알고리즘들이 제안되었다. 이 논문에서는 현재 프레임의 매크로블록과 이전프레임의 매크로블록간의 Sum of Absolute Difference를 이용하여 탐색영역을 변경하는 새로운 예측 방법을 제시한다. 실험결과에 따르면 우리가 제안한 알고리듬은 FS와 비슷한 PSNR을 유지하면서 속도가 크게 향상된 것을 볼 수 있었다.