会员体验
专利管家(专利管理)
工作空间(专利管理)
风险监控(情报监控)
数据分析(专利分析)
侵权分析(诉讼无效)
联系我们
交流群
官方交流:
QQ群: 891211   
微信请扫码    >>>
现在联系顾问~
热词
    • 2. 发明公开
    • 그래픽 가속기 기반의 고속 영상 처리 방법 및 그 장치
    • 基于图形处理单元的高速图像处理方案
    • KR1020080073511A
    • 2008-08-11
    • KR1020070012215
    • 2007-02-06
    • 삼성전자주식회사인하대학교 산학협력단
    • 원석진오윤제조성대홍태화김수균이민우박인규이만희
    • G06T1/00G06T1/20G06F3/00
    • G06T15/005H04N19/42H04N19/423H04N19/436H04N19/63
    • A high-speed image processing method based on a graphics accelerator and an apparatus thereof are provided to assign an input texture and an output texture equally in at least one FBO to use an output result as input data without an additional process, thereby shortening a process time and reducing the usage of a memory. A pixel shader(30) of a GPU(Graphic Processing Unit) receives and processes texture data(302) of 32-bit floating-point data, and has at least one FBO(Frame Buffer Object)(333). An input image, an image processing target, is loaded to a video memory as the texture data. The pixel shader performs a specific algorithm for image processing for the input image, and outputs a result to at least one FBO as the texture data. An application performs rectangular rendering matching the entire screen and performs the binding of the contents of the FBOs to a texture. Finally, the image processing result performed by the pixel shader is rendered to a viewport. Floating-point data generated in image processing is represented by assigning the type of input and output textures to the type of a 32-bit real number.
    • 提供了一种基于图形加速器及其装置的高速图像处理方法,用于在至少一个FBO中均匀地分配输入纹理和输出纹理,以使用输出结果作为输入数据,而不需要额外的处理,从而缩短处理 时间和减少内存的使用。 GPU(图形处理单元)的像素着色器(30)接收并处理32位浮点数据的纹理数据(302),并且具有至少一个FBO(帧缓冲器对象)(333)。 输入图像,图像处理目标,作为纹理数据被加载到视频存储器。 像素着色器对输入图像执行图像处理的特定算法,并将结果输出到至少一个FBO作为纹理数据。 应用程序执行与整个屏幕匹配的矩形渲染,并执行FBO内容与纹理的绑定。 最后,由像素着色器执行的图像处理结果被呈现给视口。 通过将输入和输出纹理的类型分配给32位实数的类型来表示图像处理中生成的浮点数据。
    • 5. 发明公开
    • 파노라마 사진 자동 촬영 방법
    • 自动拍摄全景照片的方法
    • KR1020090065914A
    • 2009-06-23
    • KR1020070133459
    • 2007-12-18
    • 삼성전자주식회사
    • 손병준김수균홍태화조성대오상욱
    • H04N5/262
    • H04N5/23216H04N5/232H04N5/23238
    • A method for automatically capturing a panorama photo is provided to estimate an auto-capturing time point quickly and correctly by using a motion estimation technique. An initial image is obtained by responding to a request for capturing a panorama photo(215). Movement of an imaging device capturing the initial image is checked by applying a motion estimation technique to a received image(225). A capturing direction is automatically determined based on the checked movement of the imaging device(230). The movement of the imaging device is checked by selectively applying the motion estimation technique capable of estimating a horizontal or vertical direction at first in response to the determined capturing direction(235).
    • 提供了一种自动捕获全景照片的方法,通过使用运动估计技术快速正确地估计自动捕获时间点。 通过响应拍摄全景照片的请求获得初始图像(215)。 通过对接收到的图像应用运动估计技术来检查拍摄初始图像的成像装置的移动(225)。 基于所检查的成像装置的移动来自动确定拍摄方向(230)。 通过选择性地应用能够响应于所确定的捕获方向(235)首先估计水平或垂直方向的运动估计技术来检查成像装置的移动。
    • 6. 发明授权
    • 파노라마 사진 자동 촬영 방법
    • 自动拍摄全景照片的方法
    • KR101409653B1
    • 2014-06-19
    • KR1020070133459
    • 2007-12-18
    • 삼성전자주식회사
    • 손병준김수균홍태화조성대오상욱
    • H04N5/262
    • H04N5/23216H04N5/232H04N5/23238
    • 본 발명은 파노라마 촬영 방법에 관한 것으로서, 특히 촬영 장치의 움직임을 확인하여 자동으로 파노라마 사진을 구성하는 각 화상을 촬영하는 방법에 관한 것이다. 본 발명의 파노라마 촬영방법은 파노라마 사진 촬영 방법에 있어서, (a)파노라마 사진 촬영의 요청에 대응하여 초기 영상을 획득하는 과정과, (b)현재 입력되는 영상에 모션 추정 기법을 적용하여 상기 초기 영상을 획득한 촬영 장치의 움직임을 확인하는 과정과, (c)상기 확인된 촬영장치의 움직임에 기초하여 촬영방향을 자동으로 결정하는 과정과, (d)결정된 상기 촬영방향에 대응하여, 수평 또는 수직 방향에 대해 우선적으로 움직임 추정이 가능한 모션 추정 기법을 선택적으로 적용하고 촬영 장치의 움직임을 확인하는 과정과, (e)상기 확인한 해당 촬영 장치의 움직임이 촬영 시점 결정을 위해 미리 설정된 임계값에 도달하였음을 확인함으로써 각 화상의 촬영 시점을 판단하는 과정과, (f)상기 각 화상의 촬영 시점에서 각 화상을 촬영하는 과정을 포함한다.
      파노라마, 사진, 촬영, 자동, 움직임
    • 7. 发明公开
    • 파노라마 사진 촬영 방법
    • 拍摄全景照片的方法
    • KR1020090022054A
    • 2009-03-04
    • KR1020070087087
    • 2007-08-29
    • 삼성전자주식회사
    • 김수균문재원조성대오윤제정희원
    • H04N5/262
    • H04N5/23293G03B37/02G03B37/04G06T7/223G06T7/32G06T2207/10016H04N5/23238H04N5/23245
    • A method for photographing a panorama picture is provided to recognize the motion information of a terminal at which a camera is mounted through an image processing operation when photographing a panorama picture and to determine a photographing point automatically according to the motion information. When photographing an image, the motion of a camera device is detected by comparing an inputting image with a previous image in real time based on a motion estimation technique to which an exposure calibration is applied(200-220). The distance of a real photographing region of the image by confirming the image photographing information, and a critical value is set up based on the distance(230). A photographing point of each image is determined by confirming that the motion depending on the photographing direction reaches the critical value(240). An image is inputted manually or automatically at the photographing point of each image(250).
    • 提供拍摄全景图像的方法,以便在拍摄全景图像时通过图像处理操作来识别安装相机的终端的运动信息,并根据运动信息自动确定拍摄点。 当拍摄图像时,基于施加了曝光校准的运动估计技术(200-220),通过将输入图像与先前图像实时比较来检测相机装置的运动。 通过确定图像拍摄信息的图像的实际拍摄区域的距离和临界值,基于距离(230)设置。 通过确认取决于摄影方向的运动达到临界值(240)来确定每个图像的拍摄点。 手动或自动地在每个图像的拍摄点输入图像(250)。
    • 8. 发明授权
    • 다해상도 3차원 모델 생성 방법
    • 多分辨率3D模型的生成方法
    • KR100810326B1
    • 2008-03-04
    • KR1020060098602
    • 2006-10-10
    • 삼성전자주식회사
    • 김수균
    • G06T17/00
    • G06T17/20
    • A method for generating a multi-resolution 3D(three-dimensional) model is provided to solve a problem that a multi-resolution 3D model cannot be generated and important parts are simultaneously eliminated due to that an applied multi-scale classification operator cannot accurately discriminate ridge and valley lines to find feature lines. Mesh data including texture information with respect to a 3D model is received from a 3D range scanning system(S100). The received mesh data is checked, and the first contour image is detected by applying a contour detection algorithm to a texture image of the mesh data(S110,S120). The second contour image is detected by applying the contour detection algorithm by using geometric data of the mesh data(S130,S140). Chrominance signals of each pixel included in the first and second contour images are combined to generate a single model map(S150). The model map is reversely mapped to a 3D model to extract feature lines(S160). A quadric error having a weight value assigned to the feature lines is calculated to generate a multi-resolution 3D model(S170).
    • 提供了一种用于生成多分辨率3D(三维)模型的方法,以解决多分辨率3D模型不能被产生的问题,并且由于应用的多尺度分类算子不能准确地识别而重要的部分被同时消除 脊线和谷线找到特征线。 从3D范围扫描系统接收到包括关于3D模型的纹理信息的网格数据(S100)。 检查所接收的网格数据,并通过对网格数据的纹理图像应用轮廓检测​​算法来检测第一轮廓图像(S110,S120)。 通过使用网格数据的几何数据应用轮廓检测​​算法来检测第二轮廓图像(S130,S140)。 包括在第一和第二轮廓图像中的每个像素的色度信号被组合以产生单个模型图(S150)。 将模型图反向映射到3D模型以提取特征线(S160)。 计算具有分配给特征线的权重值的二次误差,以生成多分辨率3D模型(S170)。
    • 10. 发明公开
    • 휴대 단말기 및 그 이미지 편집 방법
    • 移动终端及其图像编辑方法
    • KR1020090068576A
    • 2009-06-29
    • KR1020070136253
    • 2007-12-24
    • 삼성전자주식회사
    • 김수균손병준오상욱조성대홍태화
    • H04N5/262H04B1/40
    • H04M1/72522G06T3/4023G06T3/4038
    • A portable terminal and a method for editing an image are provided, which make the image that a user wants displayed on the portable terminal. A method for editing an image is as follows. If the image processed as the mosaic image is selected, the selected image is divided into the block image(330). The divided block image color mean value is produced. It determines whether the high position mosaic image for producing the mosaic image is selected or not. In the selected high position mosaic image, the mosaic image corresponding to the calculated color mean value is searched(350). The block image is replaced with the searched mosaic image and the mosaic image is indicated(355,360).
    • 提供便携式终端和用于编辑图像的方法,其使得用户希望在便携式终端上显示的图像。 用于编辑图像的方法如下。 如果选择了作为马赛克图像处理的图像,则将所选择的图像分割成块图像(330)。 产生分割块图像颜色平均值。 它确定是否选择用于产生马赛克图像的高位置马赛克图像。 在所选择的高位置马赛克图像中,搜索与所计算的色彩平均值相对应的马赛克图像(350)。 块图像被替换为所搜索的马赛克图像并且马赛克图像被指示(355,360)。