-
公开(公告)号：KR1020140091094A
公开(公告)日：2014-07-21
申请号：KR1020120152490
申请日：2012-12-24
Applicant: 한화테크윈 주식회사 , 고려대학교 산학협력단
CPC classification number: G06T5/003 , G06T2207/10024 , G06T2207/10028 , G06K9/00 , G06T5/00 , G06T7/00 , H04N5/217
Abstract: The present invention relates to a device and a method for processing an image and, more particularly, to a device and a method for processing an image which can restore an image when there is no fog from a fog image. The method for processing an image comprises the steps of estimating a pixel depth from an image including fog using a difference between RGB channels; refining the pixel depth by filtering the image whose pixel depth is estimated; estimating a global atmospheric source from the image whose pixel depth is refined; calculating an airlight index from the gray scale-converted image; and restoring an image without fog using the image including fog, the refined pixel depth, the global atmospheric source, and the airlight index.
Abstract translation: 本发明涉及一种用于处理图像的装置和方法，更具体地，涉及一种用于处理图像的装置和方法，该图像可以在雾图中没有雾时恢复图像。 用于处理图像的方法包括以下步骤：使用RGB信道之间的差异从包括灰雾的图像估计像素深度; 通过对像素深度估计的图像进行滤波来改善像素深度; 从像素深度精细化的图像估计全局大气源; 从灰度转换图像计算航空灯指数; 并且使用包括雾的图像，精细像素深度，全局大气源和航空灯指数来恢复图像而不起雾。
-
公开(公告)号：KR1020140080744A
公开(公告)日：2014-07-01
申请号：KR1020120146621
申请日：2012-12-14
Applicant: 한화테크윈 주식회사 , 고려대학교 산학협력단
CPC classification number: H04N5/33 , H04N5/2258 , H04N5/2351 , H04N5/243 , H04N5/265 , H04N5/332
Abstract: The present invention relates to an image fusion device and a method thereof and, more particularly, to an image fusion device and a method thereof for fusing infrared rays images and visible images by solving an error of infrared rays images photographed by an infrared (IR) camera and an error of visible images photographed by a visible camera. The image fusion method comprises the steps of: adjusting a first image of an object to be photographed as a visible image produced by the visible camera and a second image of the object as an infrared rays image produced by the IR camera; converting the brightness of the second image adjusted based on the brightness of the adjusted first image; and fusing the brightness of the converted second image by the adjusted first image.
Abstract translation: 图像融合装置及其方法技术领域本发明涉及图像融合装置及其方法，更具体地涉及一种通过解决由红外线（IR）拍摄的红外线图像的误差来对红外线图像和可见图像进行融合的图像融合装置及其方法， 相机和由可见相机拍摄的可见图像的错误。 所述图像融合方法包括以下步骤：将由所述可见照相机产生的可见图像和所述对象的第二图像作为红外线照相机产生的红外线图像调整为要拍摄的物体的第一图像; 转换基于所调整的第一图像的亮度而调整的第二图像的亮度; 以及通过所述经调整的第一图像将所转换的第二图像的亮度融合。
-
公开(公告)号：KR1020140015892A
公开(公告)日：2014-02-07
申请号：KR1020120081969
申请日：2012-07-26
Applicant: 한화테크윈 주식회사 , 고려대학교 산학협력단
CPC classification number: G06T7/35 , G06K9/46 , G06K9/6284
Abstract: The present invention relates to an image matching apparatus and a method thereof and more specifically, to an image matching apparatus and a method thereof capable of determining the success or failure of an image matching. The image matching method includes the following steps of: extracting a feature point of each of first and second images respectively photographed by a first camera and a second camera; performing a matching between the first and second images by using feature point information; estimating a transformation matrix which includes rotation information, parallel movement information, and size transformation information in the corresponding relation of a matching between a feature point of the first image and a feature point of the second image; calculating a variance of the transformation matrix; and determining the success or failure of the matching result of the first and second images depending on a variance calculation result. [Reference numerals] (AA) Start; (BB) End; (S10) Extract each feature point of a first image photographed by a first camera and a second image photographed by a second camera about a same object; (S20) Perform a matching between the first and second images by using feature point information; (S30) Estimate a transformation matrix which is a corresponding relation between the feature point of a first image and the feature point of the second image; (S40) Calculate the variance of the transformation matrix; (S50) Determine the success or failure of the matching result of the first and second images depending on the variance calculation result
Abstract translation: 图像匹配装置及其方法技术领域本发明涉及图像匹配装置及其方法，更具体地说，涉及能够确定图像匹配成功或失败的图像匹配装置及其方法。 图像匹配方法包括以下步骤：提取由第一相机和第二相机分别拍摄的第一和第二图像中的每一个的特征点; 通过使用特征点信息来执行第一和第二图像之间的匹配; 估计包括第一图像的特征点和第二图像的特征点之间的匹配的对应关系中的旋转信息，并行移动信息和大小变换信息的变换矩阵; 计算变换矩阵的方差; 以及根据方差计算结果确定第一和第二图像的匹配结果的成功或失败。 （附图标记）（AA）开始; （BB）结束; （S10）提取由第一相机拍摄的第一图像的每个特征点和由第二相机拍摄的关于相同对象的第二图像; （S20）通过使用特征点信息来执行第一和第二图像之间的匹配; （S30）估计作为第一图像的特征点与第二图像的特征点之间的对应关系的变换矩阵; （S40）计算变换矩阵的方差; （S50）根据方差计算结果确定第一和第二图像的匹配结果的成败
-
公开(公告)号：KR1020130126379A
公开(公告)日：2013-11-20
申请号：KR1020120050478
申请日：2012-05-11
Applicant: 고려대학교 산학협력단
IPC: G10L21/02
CPC classification number: G10L21/0208 , G10L15/04
Abstract: Provided are a signal processing device for detecting a speech recognition target section for speech recognition and a method thereof. The signal processing device comprises a speech extraction part extracting a first speech recognition target section by removing noise from input signals; an echo removing part extracting a second speech recognition target section by removing a section in which an echo related to the signal processing device is removed from the input signals; and a calculation part determining an overlapping section of the first speech recognition target section and the second speech recognition target section as the speech recognition target section for speech recognition. [Reference numerals] (110) Speech extraction part;(120) Echo removing part;(130) Filter part;(140) Calculation part
Abstract translation: 提供一种用于检测用于语音识别的语音识别目标部分的信号处理装置及其方法。 信号处理装置包括通过从输入信号中去除噪声来提取第一语音识别目标部分的语音提取部分; 通过去除与所述信号处理装置相关的回波从所述输入信号中去除的部分来提取第二语音识别目标部分的回波消除部分; 以及计算部分，确定第一语音识别目标部分和第二语音识别目标部分的重叠部分作为用于语音识别的语音识别目标部分。 （110）语音提取部;（120）回波消除部;（130）滤波部;（140）计算部
-
公开(公告)号：KR101316148B1
公开(公告)日：2013-10-08
申请号：KR1020120057687
申请日：2012-05-30
Applicant: 고려대학교 산학협력단
Abstract: PURPOSE: An image processing method of a vehicle and an apparatus thereof are provided to be used for an accessor-centered social network service (SNS) by combining multiple images which are photographed through multiple cameras which are mounted on the outside of the vehicle into one image. CONSTITUTION: An image acquiring unit (110) acquires multiple images which are photographed through multiple cameras. An image combining unit (140) combines multiple images into one image. The image combining unit recognizes an object which is included in multiple images or one image. The image combining unit indexes the recognized object. The image combining unit maps indexing information with the combined image. A transmitting unit (150) transmits the image to an image providing server. A control unit (170) controls the image acquiring unit, the image combining unit, and the transmitting unit. [Reference numerals] (110) Image acquiring unit; (120) Pre-processing unit; (130) Location information module; (140) Image combining unit; (150) Transmitting unit; (160) Receiving unit; (170) Control unit
Abstract translation: 目的：提供一种车辆及其装置的图像处理方法，用于将通过安装在车辆外部的多个照相机拍摄的多个图像组合成一个以存储器为中心的社交网络服务（SNS） 图片。 构成：图像获取单元（110）获取通过多个照相机拍摄的多个图像。 图像组合单元（140）将多个图像组合成一个图像。 图像合成单元识别包含在多个图像或一个图像中的对象。 图像组合单元对识别的对象进行索引。 图像组合单元将索引信息与组合图像进行映射。 发送单元（150）将图像发送到图像提供服务器。 控制单元（170）控制图像获取单元，图像合成单元和发送单元。 （附图标记）（110）图像获取单元; （120）预处理单元; （130）位置信息模块; （140）图像组合单元; （150）发射单元; （160）接收单元; （170）控制单元
-
Patent Agency Ranking
公开(公告)号：KR1020130097490A
公开(公告)日：2013-09-03
申请号：KR1020120019178
申请日：2012-02-24
Applicant: 한화테크윈 주식회사 , 고려대학교 산학협력단
Abstract: PURPOSE: A Sound information based situation recognition apparatus and method are provided to use mel-frequency cepstral coefficient (MFCC) characteristics and timbre characteristics together and to increase the recognition and classification performance of a sound event within a narrow space by the use of a hierarchical classifier. CONSTITUTION: A sound recognition apparatus extracts MFCC characteristics and timbre characteristics from an inputted sound signal (S61). The sound recognition apparatus calculates similarity between the MFCC characteristics and the timbre characteristics and standard sound characteristics which is obtained beforehand for multiple sound events including multiple voice events and non-voice events and evaluates sound characteristics (S63). The sound recognition apparatus classifies the sound signal as one among the multiple sound events with a hierarchical access method based on the calculated similarity (S65,S67). [Reference numerals] (AA) Start; (BB) End; (S61) Timbre characteristics are extracted from an inputted sound signal; (S63) The characteristics are evaluated; (S65) It is decided whether the sound signal is a sound or not; (S67) It is decided whether an event is a sound event or a non-sound event; (S69) It is decided whether the current situation is normal or not
Abstract translation: 目的：提供一种基于声音信息的情境识别装置和方法，以一起使用梅尔频率倒谱系数（MFCC）特征和音色特征，并通过使用分层级增加狭窄空间内的声音事件的识别和分类性能 分类。 构成：声音识别装置从输入的声音信号中提取MFCC特性和音色特征（S61）。 声音识别装置计算MFCC特性与预先对包括多个语音事件和非话音事件的多个声音事件而获得的音色特性和标准声音特性的相似度，并评估声音特性（S63）。 声音识别装置基于所计算出的相似度，利用分层存取方法将声音信号分类为多个声音事件之一（S65，S67）。 （附图标记）（AA）开始; （BB）结束; （S61）从输入的声音信号中提取音色特征; （S63）评价特征; （S65）判定声音是否为声音信号; （S67）决定事件是声音事件还是非声音事件; （S69）决定当前情况是否正常
-
公开(公告)号：KR1020110125431A
公开(公告)日：2011-11-21
申请号：KR1020100044953
申请日：2010-05-13
Applicant: 삼성전자주식회사 , 고려대학교 산학협력단
IPC: H04B1/40
CPC classification number: G06Q10/04
Abstract: PURPOSE: A method and device for generating a life log in portable terminal are provided to generate a life log through the situation of a user. CONSTITUTION: A sound environment recognizing unit(212) recognizes a sound environment corresponding to a sound signal inputted from a microphone among the preset sound environment. An image environment recognizing unit(214) recognizes an image environment corresponding to an image signal inputted from a camera among the preset image environments. A situation determining unit(220) determines a user situation corresponding to a sound environment recognition result and an image environment recognition result among the preset situation models. The situation determining unit records the determined user situation as a life log.
Abstract translation: 目的：提供一种用于在便携式终端中生成生命日志的方法和装置，以通过用户的情况生成生命日志。 构成：声音环境识别单元（212）识别在预设声音环境中与麦克风输入的声音信号相对应的声音环境。 图像环境识别单元（214）在预设图像环境中识别与从相机输入的图像信号相对应的图像环境。 情境确定单元（220）在预设情况模型中确定与声音环境识别结果和图像环境识别结果相对应的用户状况。 情况确定单元将所确定的用户状况记录为生命日志。
-
公开(公告)号：KR1020070070000A
公开(公告)日：2007-07-03
申请号：KR1020060020754
申请日：2006-03-06
Applicant: 고려대학교 산학협력단
IPC: G10L15/28
Abstract: A speech recognition apparatus and method using two-stage search are provided to perform speech recognition at a high speed without decreasing a speech recognition rate. A speech recognition apparatus includes a fast search unit(100), an N-best candidate generator(110), and a precise search unit(130). The fast search unit performs viterbi search by using Gaussian distributions which are smaller than a predetermined number of Gaussian distributions included in a pool on an inputted speech signal to generate a plurality of candidate words. The N-best candidate generator extracts candidate words in the order of confidence measures. The precise search unit performs viterbi search by using a number of Gaussian distributions which are greater than the predetermined number of Gaussian distribution included in the pool on the extracted candidate words to output recognized words.
Abstract translation: 提供了一种使用两级搜索的语音识别装置和方法，以高速执行语音识别，而不降低语音识别率。 语音识别装置包括快速搜索单元（100），N最佳候选生成器（110）和精确搜索单元（130）。 快速搜索单元通过使用比输入的语音信号中包含在池中的预定数量的高斯分布的高斯分布来执行维特比搜索以生成多个候选词。 N最佳候选生成器按照置信度的顺序提取候选词。 精确的搜索单元通过使用多个高斯分布来执行维特比搜索，该多个高斯分布大于包括在所提取的候选词中的池中的预定数量的高斯分布，以输出识别的字。
-
公开(公告)号：KR1020070050699A
公开(公告)日：2007-05-16
申请号：KR1020050108236
申请日：2005-11-11
Applicant: 고려대학교 산학협력단
IPC: G10L21/0208
Abstract: 잡음환경에 강인한 음성인식을 위한 에너지 특징 보상 방법 및 장치가 개시된다. 그 에너지 특징 보상 방법은 (a) 잡음이 없는 깨끗한 환경에서 수집한 음성 훈련데이터의 에너지 특징을 잡음환경의 에너지와 유사한 환경으로 변환시키는 단계; (b) 인식 음성데이터의 에너지 최소값이 소정의 목표 최소값보다 큰지 체크하여, 상기 음성데이터의 에너지 최소값이 상기 목표 최소값보다 작으면, 에너지의 최소값과 최대값의 중간값 이하만 ERN 변환식에 의해 변환시키는 Half-ERN을 수행하는 단계; 및 (c) 인식 음성데이터의 에너지 최소값이 목표 최소값보다 작지 않으면 에너지 특징을 소정의 방법으로 보상하는 단계를 포함함을 특징으로 한다.
본 발명에 의하면, 보다 다양한 환경에 더욱 강인한 에너지 특징 보상이 가능하게 되어 잡음 환경에서의 음성 인식률을 향상시킬 수 있다.-
公开(公告)号：KR1020070050694A
公开(公告)日：2007-05-16
申请号：KR1020050108226
申请日：2005-11-11
Applicant: 고려대학교 산학협력단
IPC: G10L21/0208 , G10L15/20
Abstract: 멀티채널 음성신호의 잡음제거 방법 및 장치가 개시된다. 그 잡음제거 방법은 (a) 음성신호로부터 환경에 따른 잡음 고유의 성분을 제거하는 단계; (b) 상기 잡음 고유성분이 제거된 음성신호로부터 음성과 잡음을 분리하는 단계; 및 (c) 상기 분리된 음성에 남아있는 잔여 잡음을 제거하는 후처리 단계를 포함함을 특징으로 한다.
본 발명에 의하면, 멀티채널 잡음 제거 시스템의 입력단에서 잡음을 제거함으로써 결과적으로 단일 채널 환경에서의 잡음처리 방법보다 향상된 성능을 얻을 수 있고, 전체 시스템의 성능을 향상시킬 수 있다.
-
-
-
-
-
-
-
-
-
Search Results
97
records
(took 0.03 seconds)
