Abstract:
PURPOSE: A display device and a method thereof are provided to supply a flowing animation by granting an independent animation effect for an object included in a first and a second window while a first window is switched to a second window. CONSTITUTION: A display unit (130) displays a window including an object. When a screen switch event is sensed, a control unit (120) switches a first window displayed on a screen to a second window and grants an animation effect set in the object. A storage unit (140) stores animation setting information for granting the set animation effect. The animation setting information includes information about an object animation period, an attribute, and a starting offset included in the first and the second windows. [Reference numerals] (110) Input unit; (120) Control unit; (121) Frame generation unit; (122) Buffer unit; (123) Window processing unit; (130) Display unit; (140) Storage unit
Abstract:
PURPOSE: A data processing device and a method thereof are provided to hide the memory access latency due to cache miss. CONSTITUTION: A buffer memory unit(120) stores a first input ray data for video rendering in the ray tracing method. A control unit(110) receives first shape data corresponding to the first input ray data from a cache(102). When the first shape data is not stored in the cache, the control unit can receive second shape data while the first shape data is patched to the cache from an external memory. The control unit outputs the first shape data with the first input ray data to a computation device(104).
Abstract:
PURPOSE: An image processing device is provided to increase spatial locality for reading memory when an acceleration structure is generated and searched wherein the acceleration structure is a space division acceleration structure. CONSTITUTION: A division unit(110) divides the first space of an input 3D model into a plurality of sub spaces to generate an acceleration structure for an input 3D model. A determination unit(120) sets a sub space as a child node with a higher search priority in the acceleration structure wherein the sub space has a higher possibility than the other sub spaces that the sub space includes for a moving route of a ray.
Abstract:
PURPOSE: A system and a method for 3d graphic rendering using multi-core processor are provide to process a task by dividing the task so that data is not be overlapped by using multi core processor. CONSTITUTION: A plurality of cores(220,230) comprises a scratch pad memory. A first memories(221,231) performs a control flow. A second memories(222,232) performs a loop acceleration. A shared memory(210) is connected to the plurality of cores. The plurality of cores is set dual core as a basic unit. The plurality of cores is comprised of one or more dual cores.
Abstract:
An apparatus and a method for avoiding bank conflicts are provided to prevent performance of a processor from being lowered due to bank conflicts. Writing/reading units access to one bank according to an access command. A scheduling unit changes a performing time point of the first command among access commands for indicating access in the same bank into a previous time point to change the first command into the second command in order to enable the access commands not to indicate access in the same bank. A switching unit(190) enables one of the writing/reading units to receive read data at the performing time point of the first command from one bank according to the second command in consideration of a preoccupying table. The scheduling unit comprises a table storage unit(160), a bank conflict estimator(180) and a table updating unit(170). The table storage unit stores the preoccupying table showing access commands.
Abstract:
A method for controlling a voltage of power supplied to a 3D graphics data processing device and the 3D graphics data processing device using the same are provided to reduce the voltage by using a slack time generated in processing the 3D graphics data of a frame or an object, thereby reducing the consumption of power used in processing the 3D graphics data. A method for controlling a voltage in processing a frame in a pipeline scheme comprises the following steps of: predicting the entire work load of a frame(510); determining the voltage and frequency of power supplied to a pipeline processing unit on the basis of the predicted work load(520); and determining a voltage and a frequency corresponding to a frame received next in consideration of a slack time generated after objects included in the received frame are processed by the pipeline processing unit(530).
Abstract:
대용량 얼굴 인식 장치 및 방법이 개시된다. 본 발명의 얼굴 인식 장치는 복수의 사람들을 등록하고, 등록된 사람들 각각에 대해 복수의 얼굴 영상을 저장하는 얼굴 데이터베이스; 입력 얼굴 영상에 대해 얼굴 데이터베이스를 참조하여 얼굴 검색을 수행하고 저장된 얼굴 영상들에 대해 얼굴 데이터베이스에 저장된 얼굴 영상들과 상기 입력 얼굴 영상의 유사한 정도를 나타내는 신뢰값을 계산하여 출력하는 얼굴 검색부; 신뢰값을 기반으로하여 얼굴 데이터베이스에 저장된 얼굴 영상들중 소정 수의 얼굴 영상들을 선택하여 후보자들을 결정하는 후보 선택부; 및 후보자들에 대해 각 후보에 해당하는 특징벡터와 입력 얼굴 영상의 특징벡터를 각각 비교하여 인식하는 얼굴 검증부를 포함함을 특징으로한다.
Abstract:
대용량 얼굴 인식 장치 및 방법이 개시된다. 본 발명의 얼굴 인식 장치는 복수의 사람들을 등록하고, 등록된 사람들 각각에 대해 복수의 얼굴 영상을 저장하는 얼굴 데이터베이스; 입력 얼굴 영상에 대해 얼굴 데이터베이스를 참조하여 얼굴 검색을 수행하고 저장된 얼굴 영상들에 대한 신뢰값을 계산하여 출력하는 얼굴 검색부; 신뢰값을 기반으로하여 얼굴 데이터베이스에 저장된 얼굴 영상들중 소정 수의 얼굴 영상들을 선택하여 후보자들을 결정하는 후보 선택부; 및 후보자들에 대해 각 후보에 해당하는 특징벡터와 입력 얼굴 영상의 특징벡터를 각각 비교하여 인식하는 얼굴 검증부를 포함함을 특징으로한다.
Abstract:
사람검출방법 및 장치와 이를 이용한 사생활 보호시스템이 개시된다. 사생활 보호시스템은 프레임간의 움직임정보를 이용하여 현재 프레임영상으로부터 적어도 하나 이상의 움직임영역을 검출하는 움직임영역 검출부; 상기 검출된 각 움직임영역에 대하여 사람의 형태정보를 이용하여 사람을 검출하고, 소정의 추적영역에 대하여 이전 프레임영상에서 사람으로 검출된 움직임영역을 추적하는 사람 검출 및 추적부;상기 사람으로 검출되거나 추적된 움직임영역에서 얼굴을 검출하고, 디스플레이하기 위하여 검출된 얼굴을 모자이크처리하는 모자이크처리부; 및 상기 사람으로 검출되거나 추적된 움직임영역을 검색을 위한 소정의 라벨 및 위치정보와 함께 프레임별로 저장하는 저장부를 포함한다.