Abstract:
본 발명은 중요도를 갖는 부분과 중요도가 낮은 부분과 중간부분으로 가중치를 설정하고, 시간경과에 따라서 중요도에 비례하여 업데이트 주기를 달리함으로써 중요한 부분을 더욱 정밀하게 송수신할 수 있도록 하여 화상품질을 높이도록 하는 동화상 압축/복원 방법을 제공한다. 상기 목적을 위하여 본 발명에는 화상을 공간적으로 중요도에 따라 분류하는 단계; 상기 중요도에 따라서 분류된 화상에 대하여 가중치를 부여하는 단계; 상기 가중치가 높은 화상에 대하여 시간경과에 따른 업데이트 빈도수를 높게하고, 상기 가중치가 낮은 화상에 대하여 시간경과에 따른 업데이트 빈도수를 낮게하는 화상업데이트 단계를 포함하는 것을 특징으로 하는 기술구성이 개시된다.
Abstract:
본 발명은 하드웨어적으로 얼굴을 트레킹함으로써 전력소모를 최소화하도록 한 얼굴트레킹 동화상 송수신장치를 제공한다. 상기 목적을 위하여 본 발명에는 멀티미디어 하드웨어를 이용하여 동화상을 전송시키는 동화상 송수신 장치에 있어서: 상기 멀티미디어 하드웨어는 콘트롤 프로세서; 상기 콘트롤 프로세서에 연결되어 압축시나 복원시에는 움직임을 추정하지만 얼굴 트레킹시에는 얼굴의 타원을 추적하는 모션 에스티메이션부; 상기 콘트롤 프로세서에 연결되어 압축시나 복원시에는 이산/ 역이산 코사인 변환을 수행하지만 얼굴 트레킹시에는 페이스 세부구조를 찾아주는 기능을 수행하는 이산/역이산 코사인 변환부를 포함하는 기술구성이 개시된다.
Abstract:
A motion estimating method is provided to reduce degradation of a compression rate of a video compression codec due to inaccuracy of motion estimation, and reduce the amount of calculation and hardware complexity in the motion estimating process. The first bit-plane of S2BT is created by using a mean value of pixels in a threshold window(100). A resonable local threshold is obtained by using second differentiation values of an image to create the second bit-plane(102). An SMVP(Spatial Motion Vector Prediction) scheme using edge information obtained by the second differentiation values of the image and motion vector information of peripheral MB(MicroBlocks) is used to avoid a binary matching process with a large amount of calculation(104). A best correlation between an MB of a previous frame and that of a current frame is searched by using only 1-bit boolean operators with respect to the two bit-planes as obtained(106).
Abstract:
A frame interpolating apparatus and a frame interpolating method are provided to insert an interpolated frame generated by a corrected motion vector between frames sent from an encoder to improve the quality of an output image. A frame interpolating apparatus includes a decoder(1) for receiving encoded video data and a motion vector from an encoder and decoding the encoded video data, an object segmentation module(2) for segmenting the decoded image received from the decoder by objects, and a motion vector correction module(3) for correcting the motion vector according to the decoded image received from the decoder and the motion vector. The apparatus further includes a pixel motion vector generation module(4) for generating a motion vector with respect to each pixel according to object segmentation information from the object segmentation module and the motion vector corrected by the motion vector correction module, and a motion compensated frame interpolation module(5) for generating an interpolated frame according to the motion vector with respect to each pixel.
Abstract:
본 발명은 중요도를 갖는 부분과 중요도가 낮은 부분과 중간부분으로 가중치를 설정하고, 시간경과에 따라서 중요도에 비례하여 업데이트 주기를 달리함으로써 중요한 부분을 더욱 정밀하게 송수신할 수 있도록 하여 화상품질을 높이도록 하는 동화상 압축/복원 방법을 제공한다. 상기 목적을 위하여 본 발명에는 화상을 공간적으로 중요도에 따라 분류하는 단계; 상기 중요도에 따라서 분류된 화상에 대하여 가중치를 부여하는 단계; 상기 가중치가 높은 화상에 대하여 시간경과에 따른 업데이트 빈도수를 높게하고, 상기 가중치가 낮은 화상에 대하여 시간경과에 따른 업데이트 빈도수를 낮게하는 화상업데이트 단계를 포함하는 것을 특징으로 하는 기술구성이 개시된다. STMAC, MPEG-4, 템포랄 프로세싱
Abstract:
An error concealing apparatus and a method thereof are provided to obtain a sharp image without an edge breakup between a normally received image block and a concealed image block in an unstable communication environment. A decoder(1) receives video data coded by a coder and motion vectors, and decodes them. An edge detecting module(2) detects an edge of the decoded image received from the decoder. An object detecting module(3) detects an object through the decoded image and the detected edge received from the decoder. An error concealing module(4) variably selects neighbor blocks by using the object information obtained from the edge detecting module and the object detecting module, and performs block matching.
Abstract:
A video frame re-compressing method in which a down-sampling type and a maximum-minimum quantization type are combined and an apparatus therefor are provided to combine the down-sampling type and the maximum-minimum quantization type which are relatively simple without using a frequency conversion type in which a hardware size is greatly occupied and a processing time is required for a long time. A re-compressor includes a down-sampling unit(10), an analyzing unit(20), a quantization unit(30), and an integrating unit(40). The down-sampling unit(10) compresses a unit block in horizontal and vertical two directions. The analyzing unit(20) analyzes a result outputted from the down-sampling unit(10), and selects pixels down-sampled in the direction in which difference between the maximum value and the minimum value is greatly generated between the two directions. The quantization unit(30) quantizes and re-compresses the respective pixels down-sampled in the direction selected by the analyzing unit(20). The integrating unit(40) integrates the re-compression data quantized in the quantization unit(30) and a header for inverse compression, and transmits the integrated data to a memory.
Abstract:
A fluctuation correcting method in a photographing device is provided to reduce quantity of calculation if the photographing device is shaken by the fluctuation or the like, thereby correcting the fluctuation in real time and improving quality of an image. A fluctuation correcting method in a photographing device comprises the following steps of: selecting a sub image(100); generating an LMV(Local Motion Vector)(104); obtaining a GMV(Global Motion Vector) by performing global motion estimation by using the LMVs each obtained at sub image areas(106); and determining whether the GMV, obtained in the above second step, is motion intended by a user, and performing global motion correction for correcting an image if the motion is not intended motion(108).