Abstract:
PURPOSE: An image processing method and an apparatus thereof through a repetition pattern are provided to efficiently match an image including a plurality of repetition patterns and to efficiently recognize an image including a repetition pattern. CONSTITUTION: A feature extracting unit(120) extracts a plurality of feature points from an inputted image. A classifying unit(130) classifies a plurality of feature points into one or more first feature points and one or more second feature points extracted from a repeated pattern. The classifying unit generates a classifier about each repetition pattern by using the first feature points. [Reference numerals] (110) Inputting unit; (120) Feature extracting unit; (130) Classifying unit; (140) Matching unit; (150) Image DB
Abstract:
PURPOSE: An image conversion apparatus and an image conversion method are provided to offer clear gray image about the same color values. CONSTITUTION: A color space converter(122) converts the color component into a color space including a brightness component. A contrast generator(124) creates color contrast values by using the difference of chroma component and the brightness component. Color contrast values are created by using the difference of the brightness component and the chroma component among the pixels which are converted by the color space converter. A gray image generator(126) creates a grey image by using the created color contrast values.
Abstract:
PURPOSE: A coding establishing apparatus, an image multiplex encoding method and a recording media recorded with a computer for increasing coding efficiency of the image are provided to supply a coding interval setting device about an image. CONSTITUTION: A scaler(310) reduces the resolution of an image. An encoder(320) encodes the image in which resolution is reduced and the coding result information about the reduced image as described above is created. A corrector(330) revises the coding result information and the coding result information about image is created. The encoded coding result of image satisfies the set goal coding result.
Abstract:
본발명은, 비디오영상에서사람과배경을분리하기위한방법에있어서, 제1 프레임에서의사람영역과배경영역을추정하는단계; 상기추정된사람영역과배경영역각각에대한가우시안혼합모델을생성하는단계; 상기사람영역에대한가우시안혼합모델과상기배경영역에대한가우시안혼합모델을합쳐전체영상에대한가우시안모델을생성하고, 제2 프레임의정보를이용하여상기전체영상에대한가우시안혼합모델을갱신하는제1 갱신단계; 상기갱신된전체영상에대한가우시안혼합모델을이용하여에너지함수생성하고, 생성된에너지함수의최소화를통해상기제2 프레임의사람영역을분리하는단계; 를포함할수 있다.
Abstract:
본 발명은 부호화 모드 결정 장치, 영상 부호화 방법 및 장치와 그를 위한 컴퓨터로 읽을 수 있는 기록매체에 관한 것이다. 본 발명은 영상 신호를 디스플레이 신호로 변환하고 변환된 디스플레이 신호를 이용하여 부호화 비용을 계산하여 부호화 비용에 따라 부호화 모드를 결정하는 것을 특징으로 하는 부호화 모드 결정 장치를 제공한다. 본 발명에 의하면, 디스플레이 단에서 표시되지 않는 정보들을 전송하지 않게 함으로써 적은 전송량만으로도 디스플레이 단에서 동일한 화질을 얻을 수 있으며, 그를 통해 영상의 압축 성능을 향상시킬 뿐만 아니라 더욱 좋은 영상을 모바일 기기에서 표시할 수 있다. 영상, 부호화, 모드, 율-왜곡 비용, YUV, RGB 565
Abstract:
PURPOSE: Image noise removal using a non-local average filtering method in a frequency conversion domain is provided to set adaptive variables to each sub-band in the frequency conversion domain by using the non-local average filtering method, thereby improving visual quality. CONSTITUTION: A noise image is converted into a frequency conversion domain by a frequency conversion(11). Conversion variables of each sub-band(12) are modeled with generalized Gaussian distribution(14). A bandwidth of a non-local average filter is estimated in the conversion variables by estimating local noise statistics(15). The bandwidth is applied to non-local average filtering(16). The conversion variables are converted into a noise removal image through inverse conversion(17). [Reference numerals] (1) Hypothesis verification in each pixel; (11) Frequency conversion; (12) Sub-band 1; (14) Sub-band noise statistical property estimation; (16) Non-local average filtering; (17) Inverse conversion; (2) Local noise statistical property estimation; (3) Local adaptive bandwidth estimation; (AA) Noise image; (BB) Sub-band K; (CC) Noise removing image
Abstract:
개시된 기술에 따른 영상 처리 방법은 (a) 입력된 이미지로부터 복수의 특징점들을 추출하는 단계, (b) 상기 복수의 특징점들을 반복적인 패턴들로부터 추출된 적어도 하나의 제1 특징점과 두드러진 적어도 일부로부터 추출된 적어도 하나의 제2 특징점으로 분류하는 단계 및 (c) 상기 제1 특징점들을 이용하여 상기 반복적인 패턴들 각각에 대한 분류기를 생성하는 단계를 포함한다.