公开(公告)号：JP2009268021A

公开(公告)日：2009-11-12

申请号：JP2008118397

申请日：2008-04-30

Applicant: Sony Corp ,  ソニー株式会社

Inventor： YAMADA SHINOBU ,  MORIMURA TAKUO ,  NAGANO KOSUKE ,  MORI HIDEKI ,  FUKAZAWA KENTARO ,  AOKI HIRONARI ,  HARASAKI SHUNSUKE ,  KAN YASUHIKO ,  INOMATA SEIICHI ,  KONDO TETSUJIRO

IPC: H04N7/01 ,  G06T3/40

Abstract: PROBLEM TO BE SOLVED: To predict an image with higher image quality by acquiring information indicative of a continuous region which is present around an attentional pixel. SOLUTION: A prediction tap extraction section 61 extracts from first image data a prediction tap used for predictive calculation for predicting the attentional pixel of second image data, a matching position acquisition section 65 acquires a matching position representing a position at which a matching block most similar to a reference block is present from a matching range representing a range in which a plurality of pixels constituting a first image are present, and a direction class sorting section 66 sorts the pixel of interest into one of a plurality of direction classes on the basis of the matching position. A coefficient ROM 67 outputs a prediction coefficient corresponding to the direction class of the pixel of interest from among prediction coefficients calculated beforehand by learning, and a predictive calculation section 68 predicts the attentional pixel in the second image data by predictive calculation using the output prediction coefficient and a plurality of pixels constituting the extracted prediction tap. COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：通过获取指示出现在注意像素周围的连续区域的信息来预测具有更高图像质量的图像。 解决方案：预测抽头提取部分61从第一图像数据中提取用于预测第二图像数据的注意像素的预测计算的预测抽头，匹配位置获取部分65获取表示匹配的位置的匹配位置 从表示其中存在构成第一图像的多个像素的范围的匹配范围存在与参考块最相似的块，并且方向类别排序部66将感兴趣的像素分成多个方向类中的一个， 匹配位置的基础。 系数ROM67从通过学习预先计算的预测系数中输出与感兴趣像素的方向类别相对应的预测系数，并且预测计算部68通过使用输出预测系数的预测计算来预测第二图像数据中的注意像素 以及构成所提取的预测抽头的多个像素。 版权所有（C）2010，JPO＆INPIT

公开(公告)号：JP2003235039A

公开(公告)日：2003-08-22

申请号：JP2002033969

申请日：2002-02-12

Applicant: SONY CORP

Inventor： KONDO TETSUJIRO ,  MUKAI HITOSHI ,  ISHIKAWA TAKANORI ,  OKUMURA AKIHIRO ,  OTSUKI TOMOYUKI ,  TANAKA TAKESHI ,  YAMADA SHINOBU

IPC: H04N11/04 ,  H03M7/40 ,  H04N7/24 ,  H04N19/00 ,  H04N19/126 ,  H04N19/136 ,  H04N19/176 ,  H04N19/186 ,  H04N19/196 ,  H04N19/98

Abstract: PROBLEM TO BE SOLVED: To provide an apparatus and method for realizing improvement of a compression rate and enhancement of image quality in coding of data with a plurality of signal components. SOLUTION: The correlation of signal components of data having a plurality of the signal components is obtained and coded data sharing a quantization code, or a minimum value, or a dynamic range of the signal component discriminated of presence of the correlation are produced. The propriety of sharing is discriminated by verifying S/N as a difference between a decoded pixel value and an original pixel value in the case of sharing the quantization code, or comparing quantization codes/bit patterns of each signal component, or comparing the standard deviation and the dynamic range of each signal component or comparing the standard deviation and the average, the sharing of the minimum value or the dynamic range or the quantization code is performed only for the components with high correlation. COPYRIGHT: (C)2003,JPO

公开(公告)号：JP2009253873A

公开(公告)日：2009-10-29

申请号：JP2008102105

申请日：2008-04-10

Applicant: Sony Corp ,  ソニー株式会社

Inventor： AOKI HIRONARI ,  MORIMURA TAKUO ,  YAMADA SHINOBU ,  NAGANO KOSUKE ,  FUKAZAWA KENTARO ,  MORI HIDEKI ,  KAN YASUHIKO ,  HARASAKI SHUNSUKE ,  INOMATA SEIICHI ,  KONDO TETSUJIRO

IPC: H04N7/01 ,  G06T3/40 ,  H04N1/387

Abstract: PROBLEM TO BE SOLVED: To generate an image signal having high resolution higher than that in the prior art, and also generate an image signal in which linking is reduced, when converting an image from an input image to an output image. SOLUTION: A feature value class classifying part 233 computes a position relation between a pixel of interest among a group of pixels forming the horizontal waveform of the input image and an edge region included in the waveform as a phase in order to convert the image from the input image to the output image, and classifies the pixel of interest based on the phase. A sum of product operation part 240 predictively operates pixels forming the output image by using a prediction coefficient specified by the feature value of the pixel of interest classified by the feature value class classifying part 233 and a group of data of pixels including the pixel of interest in the input image. This invention can be applied for example to an image conversion device. COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：为了生成比现有技术高的分辨率的图像信号，并且当将图像从输入图像转换为输出图像时，还生成缩小链接的图像信号。 解决方案：特征值类分类部分233计算形成输入图像的水平波形的像素组和波形中包括的边缘区域之间的关注像素之间的关系像素之间的位置关系，以便将 从输入图像到输出图像的图像，并且基于相位对感兴趣的像素进行分类。 产品操作部240的和通过使用由特征值分类部分233分类的感兴趣像素的特征值指定的预测系数和包括感兴趣像素的像素的数据组来预测地操作形成输出图像的像素 在输入图像中。 本发明可以应用于例如图像转换装置。 版权所有（C）2010，JPO＆INPIT

公开(公告)号：JP2009252021A

公开(公告)日：2009-10-29

申请号：JP2008100452

申请日：2008-04-08

Applicant: Sony Corp ,  ソニー株式会社

Inventor： AOKI HIRONARI ,  MORIMURA TAKUO ,  YAMADA SHINOBU ,  NAGANO KOSUKE ,  FUKAZAWA KENTARO ,  MORI HIDEKI ,  KAN YASUHIKO ,  HARASAKI SHUNSUKE ,  INOMATA SEIICHI ,  KONDO TETSUJIRO

IPC: G06T3/40

Abstract: PROBLEM TO BE SOLVED: To generate an image signal whose slanting line is smoothly reproduced when IP (interlace progressive) conversion is applied to it. SOLUTION: An input part 81 inputs an image, as an input image, whose noted line is thinned out. A class classification part 84 classifies each picture element that composes a noted line in an image, as an output image, where the thinning of a noted line is not made. An identical class determination part 86 computes intervals M and N showing a positional relationship between a set of picture element groups classified into a class identical with noted picture elements of a noted line and the noted picture elements. A class sorting part 87 sorts noted picture elements into classes based on the intervals M and N. A product-sum operation part 91 performs forecasting calculation on the noted picture elements using forecasting coefficients identified based on respective classes of noted picture elements classified by the class classification part 84 and class sorting part 87 and a forecasting tap in the input images having a predefined positional relationship with the noted picture elements, thus outputting an output image as a forecasting calculation result. The invention is applicable to image conversion equipment. COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：为了生成当应用IP（隔行逐行）转换时，其倾斜线平滑地再现的图像信号。

解决方案：输入部分81输入作为输入图像的图像，其注释线被稀疏化。 类分类部分84将构成图像中的所述线的每个图像元素分类为输出图像，其中未进行所述线的变薄。 相同的类确定部件86计算间隔M和N，该间隔M和N表示分类为与所述线条的所述图像元素相同的类的一组图像元素组与所述图像元素之间的位置关系。 类别排序部分87基于间隔M和N将所标注的图像元素分类成类。乘积和运算部分91使用基于由类别分类的所标记的图像元素的各个类别所识别的预测系数来对所述图像元素进行预测计算 分类部分84和分类部分87以及与所述图像元素具有预定位置关系的输入图像中的预测抽头，从而输出输出图像作为预测计算结果。 本发明适用于图像转换设备。 版权所有（C）2010，JPO＆INPIT

公开(公告)号：JP2009182936A

公开(公告)日：2009-08-13

申请号：JP2008022710

申请日：2008-02-01

Applicant: Sony Corp ,  ソニー株式会社

Inventor： FUKAZAWA KENTARO ,  MORIMURA TAKUO ,  YAMADA SHINOBU ,  NAGANO KOSUKE ,  MORI HIDEKI ,  AOKI HIRONARI ,  SUGA YASUHIKO ,  HARASAKI SHUNSUKE ,  INOMATA SEIICHI ,  KONDO TETSUJIRO

IPC: H04N7/01 ,  G06T3/40

Abstract: PROBLEM TO BE SOLVED: To predict an image of higher image quality by acquiring a waveform pattern in a class tap accurately. SOLUTION: A prediction tap extracting portion 31 extracts a prediction tap to be used for a prediction operation to predict a remarkable pixel which is a pixel to be noticed in a second image data, an adjacent waveform classifying portion 33 classifies the remarkable pixel as a first class based on a plurality of pixels constituting an adjacent waveform class tap extracted by an adjacent tap extracting portion 32, a wide-band waveform class classifying portion 35 classifies the remarkable pixel as a second class based on a plurality of pixels constituting a wide-band waveform class tap extracted by a wide-band tap extracting portion 34, a coefficient ROM 36 outputs a prediction coefficient corresponding to a combination of the first class and the second class, and a prediction operating portion 37 calculates the remarkable pixel of the second image data by the prediction operation which uses the output prediction coefficient and a plurality of pixels constituting the prediction tap. This invention is applied to a television receiving set, for example. COPYRIGHT: (C)2009,JPO&INPIT

Abstract translation: 要解决的问题：通过精确地获取类别抽头中的波形图案来预测更高图像质量的图像。 解决方案：预测抽头提取部分31提取用于预测操作的预测抽头以预测作为第二图像数据中要注意的像素的显着像素，相邻波形分类部分33将显着像素 作为基于由相邻抽头提取部分32提取的构成相邻波形类别抽头的多个像素的第一类，宽带波形类分类部分35基于构成第一类的多个像素将该显着像素分类为第二类 由宽带抽头提取部分34提取的宽带波形分类抽头，系数ROM36​​输出与第一类和第二类的组合相对应的预测系数，并且预测运算部37计算出 通过使用输出预测系数的预测操作和构成预测抽头的多个像素的第二图像数据。 本发明例如适用于电视接收机。 版权所有（C）2009，JPO＆INPIT

公开(公告)号：JP2009278554A

公开(公告)日：2009-11-26

申请号：JP2008129942

申请日：2008-05-16

Applicant: Sony Corp ,  ソニー株式会社

Inventor： MORIMURA TAKUO ,  YAMADA SHINOBU ,  NAGANO KOSUKE ,  FUKAZAWA KENTARO ,  MORI HIDEKI ,  AOKI HIRONARI ,  INOMATA SEIICHI ,  KAN YASUHIKO ,  HARASAKI SHUNSUKE ,  KONDO TETSUJIRO

IPC: H04N7/01 ,  G06T3/40

Abstract: PROBLEM TO BE SOLVED: To generate more beautiful and natural image. SOLUTION: An area extraction unit 111 extracts a class tap from an input SD signal, and a pattern detection unit 112 and a class code determination unit 113 determine a class code. A waveform characteristic quantity calculation unit 117 generates a vector whose elements take respective values obtained based on the differences between respective luminance values of the pixels that constitute the supplied class tap and respective luminance values, which are stored beforehand as values of a plurality of kinds for every class code and are of the pixels that constitute the class tap. A prediction coefficient generation unit 118 generates weight using the generated vector and weights the coefficients, which are stored beforehand as values of the plurality of kinds for every class code and correspond to respective pixels that constitute the prediction tap, to generate final prediction coefficients. A prediction calculation unit 116 generates an HD signal by calculating a linear first-order expression in which the prediction coefficients supplied from the prediction coefficient generation unit 118 are used for the prediction tap extracted by the area extraction unit 115. COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：产生更美丽自然的图像。 解决方案：区域提取单元111从输入的SD信号中提取类别抽头，并且模式检测单元112和类别代码确定单元113确定类代码。 波形特征量计算单元117生成一个向量，该元素取各自根据构成所提供的类别抽头的像素的各个亮度值和各个亮度值之间的差值获得的各个值，这些值预先存储为多种 每个类代码都是构成类别抽头的像素。 预测系数生成单元118使用所生成的矢量来生成加权，并对作为各种代码的多种的值预先存储的系数进行加权，并对应于构成预测抽头的各个像素，生成最终的预测系数。 预测计算单元116通过计算从预测系数生成单元118提供的预测系数用于由区域提取单元115提取的预测抽头的线性一阶表达式来生成HD信号。COPYRIGHT :(（ C）2010，JPO＆INPIT

公开(公告)号：JP2009260905A

公开(公告)日：2009-11-05

申请号：JP2008160295

申请日：2008-06-19

Applicant: Sony Corp ,  ソニー株式会社

Inventor： MORI HIDEKI ,  KONDO TETSUJIRO ,  MORIMURA TAKUO ,  YAMADA SHINOBU ,  NAGANO KOSUKE ,  FUKAZAWA KENTARO ,  AOKI HIRONARI ,  INOMATA SEIICHI ,  KAN YASUHIKO ,  HARASAKI SHUNSUKE

IPC: H04N7/01 ,  G06T3/40

Abstract: PROBLEM TO BE SOLVED: To process adaptively even edited material, to obtain a statistically certain processing result as a processing result of the processing, and to reduce a user's manipulated variable or to suppress rapture or unnatural emphasis as a result. SOLUTION: A learning unit 21 and an output image generating unit 22 perform input energy classification as follows. Namely, such classification is performed that, when sharpness of a waveform in an edge part of an image is defined as a feature amount, a plurality of classes are provided in accordance with degrees of the feature amount, each of pixels constituting data Sa, Sb, Sc of SD or data I is defined as an attention pixels, and the attention pixel is sorted into any one of the plurality of classes on the basis of the feature amount. The present invention may be applicable to a system for applying resolution improving processing to edited material, for example. COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：为了处理自适应地编辑的材料，获得作为处理的处理结果的统计上特定的处理结果，并且因此减少用户的操纵变量或抑制被破坏或不自然的强调。 解决方案：学习单元21和输出图像生成单元22如下进行输入能量分类。 即，进行这样的分类：当将图像的边缘部分中的波形的锐度定义为特征量时，根据特征量的程度提供多个等级，构成数据Sa，Sb的每个像素 ，将SD或数据I的Sc定义为注意像素，并且基于特征量将注意像素分类为多个类别中的任一个。 例如，本发明可以应用于对编辑材料进行分辨率改善处理的系统。 版权所有（C）2010，JPO＆INPIT

公开(公告)号：JP2009200914A

公开(公告)日：2009-09-03

申请号：JP2008041431

申请日：2008-02-22

Applicant: Sony Corp ,  ソニー株式会社

Inventor： FUKAZAWA KENTARO ,  HAMAMATSU TOSHIHIKO ,  AOKI HIRONARI ,  MORIMURA TAKUO ,  YAMADA SHINOBU ,  NAGANO KOSUKE ,  MORI HIDEKI ,  SUGA YASUHIKO ,  HARASAKI SHUNSUKE ,  INOMATA SEIICHI ,  KONDO TETSUJIRO

IPC: H04N7/01 ,  G06T3/40 ,  G06T5/20 ,  H04N11/20

Abstract: PROBLEM TO BE SOLVED: To provide an image processing apparatus, image processing method, and program for generating an image with higher image quality. SOLUTION: A band determination units 12 1 to 12 3 determines signal bands of a plurality of component signals constituting first image data. Based on a determination result of the signal bands, a classification unit 13 determines a correlation among the plurality of component signals. Based on a combination of the determination result of the signal bands of the plurality of component signals and the correlation among the plurality of component signals, as a tap to be used for predictive operation for predicting a pixel of interest that is a pixel to which attention is payed in second image data, a tap shape in extracting a plurality of pixels, including corresponding pixels of the first image data, corresponding to the pixel of interest and being present around the corresponding pixel, from the plurality of component signals is determined for each of the plurality of component signals and the tap is used for classifying the pixel of interest. COPYRIGHT: (C)2009,JPO&INPIT

Abstract translation: 要解决的问题：提供一种用于生成具有更高图像质量的图像的图像处理装置，图像处理方法和程序。 解决方案：频带确定单元12 至12 3 确定构成第一图像数据的多个分量信号的信号频带。 基于信号频带的确定结果，分类单元13确定多个分量信号之间的相关性。 基于多个分量信号的信号频带的确定结果和多个分量信号之间的相关性的组合，作为用于预测作为注意的像素的关注像素的预测操作的抽头 在第二图像数据中支付，对于每个分量信号，确定从多个分量信号中提取包括与感兴趣的像素相对应并且存在于相应像素周围的并存在于相应像素周围的第一图像数据的对应像素的多个像素的抽头形状 的多个分量信号和分接头用于对感兴趣的像素进行分类。 版权所有（C）2009，JPO＆INPIT

公开(公告)号：JP2009224848A

公开(公告)日：2009-10-01

申请号：JP2008064220

申请日：2008-03-13

Applicant: Sony Corp ,  ソニー株式会社

Inventor： FUKAZAWA KENTARO ,  MORIMURA TAKUO ,  YAMADA SHINOBU ,  NAGANO KOSUKE ,  MORI HIDEKI ,  AOKI HIRONARI ,  KAN YASUHIKO ,  HARASAKI SHUNSUKE ,  INOMATA SEIICHI ,  KONDO TETSUJIRO

IPC: H04N7/01 ,  G06T3/40 ,  H04N1/387

CPC classification number: G06T5/001 ,  G06T5/20 ,  G06T2207/20012

Abstract: PROBLEM TO BE SOLVED: To predict an image having higher quality. SOLUTION: A prediction tap extracting section 30 extracts a prediction tap and a level limit class tap, a peak level direction class tap extracting section 31 extracts a peak level direction class tap, and a waveform class tap extracting section 32 extracts a waveform class tap, respectively. A level component class classifying section 33 classifies a target pixel into any one of a plurality of level component classes based on the peak level direction class tap and the level limit class tap. The waveform pattern class classifying section 34 classifies the target pixel into any one of the plurality of waveform pattern classes based on the waveform class tap. A coefficient ROM 35 outputs a prediction coefficient corresponding to a combination of the level component class of the target pixel and the waveform pattern class of the target pixel. A prediction calculating section 36 predicts the target pixel based on prediction calculation of the prediction coefficient and the prediction tap. COPYRIGHT: (C)2010,JPO&INPIT

Abstract translation: 要解决的问题：预测具有较高质量的图像。 解决方案：预测抽头提取部分30提取预测抽头和电平限制类别抽头，峰值电平方向类别抽头提取部分31提取峰值电平方向类别抽头，并且波形类别抽头提取部分32提取波形 分类龙头。 电平分量类分类部分33基于峰值电平方向类别抽头和电平限制类别抽头将目标像素分类为多个电平分量类别中的任何一个。 波形图案类别分类部分34基于波形类别抽头将目标像素分类为多个波形图样类别中的任何一个。 系数ROM35输出与目标像素的电平分量类别和目标像素的波形图案类别的组合相对应的预测系数。 预测计算部36基于预测系数和预测抽头的预测计算来预测目标像素。 版权所有（C）2010，JPO＆INPIT

公开(公告)号：JP2002359821A

公开(公告)日：2002-12-13

申请号：JP2001165529

申请日：2001-05-31

Applicant: SONY CORP

Inventor： KONDO TETSUJIRO ,  YAMADA SHINOBU ,  NAGANO TAKAHIRO ,  YASUOKA TOMOHIRO

IPC: H04N7/01

Abstract: PROBLEM TO BE SOLVED: To economize the capacity of a memory storing coefficient class data. SOLUTION: A class is detected from pixel data of a plurality of taps selectively fetched from an SD signal to obtain the class code CL of the target pixel of an HD signal. A memory bank 135 stores coefficient class data of each class. A coefficient generation circuit 136 generates coefficient data Wi (i=1 to n) of an estimated equation, corresponding to each estimated tap position in each class with a generation expression, including parameters (x) and (y) designating the position of an estimated tap and image quality adjustment parameters (h) and (v), on the basis of the coefficient class data of each class the parameters (h) and (v) through user operation and stores the coefficient data Wi in a memory 134. An arithmetic circuit 127 calculates the pixel data of the target pixel of the HD signal from pixel data xi of a plurality of estimated taps selectively fetched from the SD signal and the coefficient data Wi, corresponding to the code CL from the memory 134 by using the estimated equation.