Jinwuk Seok

Self-organizing map with time-invariant learning rate and its exponential stability analysis

Abstract In this paper a new self-organizing map (SOM) is developed that is suitable for digital hardware implementation. In the proposed neural model, a time-invariant learning rate is used, whereas the original Kohonen SOM uses a time-varying learning rate. There is a binary re-inforcement term in order to compensate for the lowered learning ability due to the constant learning rate. The proposed SOM is exponentially stable. The experimental results conducted with two different types of data show that the proposed method has better learning ability than the original SOM.

An efficient inter prediction mode decision method for fast motion estimation in HEVC

High Efficiency Video Coding (HEVC) has been developed as the newest video coding standard which achieves the goal of 50% bit-rate reduction with similar video quality compared to the previous standard, H.264/AVC. In order to achieve this improved coding performance, HEVC adopts various advanced techniques, which require heavy encoding complexity. Similar to the previous video coding standards, HEVC has the most computational complexity to perform motion estimation of inter prediction process. Proposed method can select one inter prediction mode and perform the motion estimation only on the selected inter prediction mode. Experimental results show that the prosed method reduces the computational complexity while minimizing the coding loss.

Fast block mode decision algorithm in H.264/AVC using a filter bank of Kalman filters for high definition encoding

In this paper, we propose a fast mode decision algorithm using a filter bank of Kalman filters for H.264/ AVC. For the highest coding efficiency in H.264/AVC, a macroblock can be coded with seven different block sizes for motion compensation in an inter mode and various spatial prediction modes in an intra mode. The conventional encoder employs a complex technique for mode decision based on a rate-distortion (RD) cost of all possible modes. Hence, for the purpose of selecting the best block mode with the minimum RD cost, the conventional procedure requires much computational burden and a very complex encoding structure. In order to reduce the complexity, we propose a fast algorithm for mode decision based on Kalman filtering to estimate RD cost of a specific block mode. Furthermore, we propose an optimized structure of H.264/AVC encoder to implement the proposed algorithm. Without considerable performance degradation, using SIMD technology, the computer simulation shows that the proposed methods are dramatically faster than the original JM 9.6 encoder.

The differential structure of sub pixels interpolated from integer pixels using n-tab FIR filters for high definition H.264 video encoding

In this paper, we propose a differential structure of subpixels which are interpolated from integer pixels using ntab linear FIR filters. The generation and management of sub-pixels are a very serious problem in high-definition H.264 video encoding. Since the amount of sub-pixels in H.264 encoding is about 16 times larger than that of integer pixels, H.264 encoder needs very high speed memory access for realtime encoding of high-definition video. In order to avoid the sub-pixel problem, we estimate the distortion value for a motion estimation of sub-pixel resolution by using the nonlinear combination of the distortion value derived from a motion estimation of related integer pixel, based on the proposed differential structure. Simulation results show that the proposed mathematical properties are valid by the sub-pixel motion estimation based on the proposed mathematical analysis. It also shows that the proposed sub-pixel motion estimation is so effective that the encoding speed is dramatically increased without a considerable depression of video quality.

ROUTE/DASH server system development for realtime UHD broadcasting

Recently, digital broadcasting has been increasingly demanded for high-definition broadcasting services such as HDR UHDTV. In addition, consumption environment of multimedia is also changing to a tendency to enjoy through various types of terminals. In order to satisfy such changes, we developed a ROUTE/DASH Server system to be inputted real-time encoded UHD stream, packetize it into ROUTE/DASH and output it to UDP, and we also verify it through developed ROUTE/DASH client. The developed ROUTE/DASH Server system provides a way to solve that the limitations of various network and receiving environment of existing MPEG-2 TS system. Using the developed system, it is expected that it can be used not only for multimedia service through broadband but also for transmission system of broadcasting system of ATSC 3.0 in which next standard broadcasting is being standardized.

Tile-level and Frame-level Parallel Encoding for HEVC

고화질 비디오에 대한 시장의 요구가 높아짐에 따라 고화질 비디오를 기존 보다 낮은 데이터 량으로 압축할 수 있는 새로운 비디오 부호화 표준 기술인 HEVC(High Efficiency Video Coding)가 최근에 개발 완료되었다. 즉 HEVC로 압축한 데이터의 양은 기존 비디오 부호화 표준인 AVC/H.264로 압축한 데이터의 양의 동일 화질 대비 약 50%로 보고되고 있다. 압축 성능이 개선된 새로운 부호화 표준에 대한 시장의 관심은 뜨겁지만 시장에 바로 활용되기 위하여서는 응용 서비스에서 요구하는 수준의 처리속도를 만족시켜야 한다. 다수의 코어가 탑재된 컴퓨터 시스템이 널리 보급된 오늘날의 개발 환경에서 부호화 처리 속도를 개선시키기 위해서 여러 각도의 병렬 부호화 적용이 필수이다. 본 논문에서는 HEVC 부호화기에 화면 분할 병렬화와 프레임 수준의 병렬화를 조합하여 적용할 때 코딩 효율 대비 병렬화로 인한 더 높은 속도 향상 결과를 가져올 수 있는 방법을 제안하였다. 즉 시스템 자원과 병렬로 처리할 프레임에 따라 화면을 적응적으로 분할하게 함으로써 코딩 효율 대비 속도 향상을 개선시킬 수 있었다. 한 화면 안에서는 다수의 타일(Tile) 단위로 병렬처리하고, 참조되지 않는 프레임(Frame)들을 병렬로 부호화하도록 본 논문을 통해 구현하였으며 Full-HD 및 4K UHD 영상을 이용하여 제안하는 방법이 코딩 효율 대비 병렬화로 인한 속도 향상이 개선되었음을 보여주었다.

Hyperbolic quotient feature map for competitive learning neural networks

In this paper, we present a new learning method called hyperbolic quotient feature map for competitive learning neural networks. The previous neural network learning algorithms didn’t consider their topological properties, and thus their dynamics were not clearly defined. We show that the weight vectors obtained by competitive learning decompose the input vector space and map it to the quotient space X/R. In addition, we define a quotient function which maps [1, ∞) ⊂ Rn to [0,1) and induce the proposed algorithm from the performance measure with the quotient function. Experimental results for pattern recognition of remote sensing data indicate the superiority of the proposed algorithm in comparison to the conventional competitive learning methods.