{"title":"Detección de regiones de interés visual mediante el Análisis Espectral Local de la imagen","authors":"J. Casanova, Germán Fernández Titos","doi":"10.1174/021435501753635596","DOIUrl":null,"url":null,"abstract":"espanolLa nocion de mapa, entendida como correspondencia biunivoca entre dos niveles de representacion en el Sistema Visual (SV), nos lleva a plantear la cuestion de como puede utilizar la vision estos mapas (que incluyen las representaciones distribuidas) para procesar la informacion de la imagen. Para ello, fijamos dos objetivos. Por un lado, analizar la retina cortical, un mapa ubicado en el cortex estriado, que sirve de soporte a una importante representacion conjunta espacio-espectral. Para generar esta representacion concebimos un modelo, simplificado, compuesto por cuatro laminas de frecuencia espacial y cuatro canales de orientacion, el cual produce dicha representacion mediante Analisis espectral local. Por otro lado, proponemos otro modelo que permite la deteccion automatica de regiones de interes visual. Este modelo se fundamenta en la hipotesis de que las respuestas a los filtros lineales locales aplicados a la imagen, y almacenadas en la retina cortical, enalan las regiones que mas destacan en la imagen acromatica, produciendo, asi, un grupo perceptivo. Complementariamente, sugerimos que aquellas regiones que `perduran' en las diferentes escalas coinciden con las de mayor saliencia visual EnglishWe understand the term map as one-to-one correspondence between two representation levels in the Visual System (VS). This notion lead us to outline the question about how the (Visual System) VS can use such map (which include distributed representations) for the image processing. In order to achieve this goal we fix two objectives. On the one, to analyse the cortical retinae, a map located in the striate cortex that contains an important joint space/spectral representation, we have elaborated a simplified model to obtain this joint representation, which is composed by layers of cells tuned to different spatial frequencies and for columns tuned to different orienttions. The joint representation is generated by applying Local Spectral Analysis to an image. On the other, we proposed a new model that allow us to detect, automatically, interesting regions in the image. It is based in the following hypothesis: the outputs of linear and local filters applied to the image, and stored in the cortical retinae, indicate the more highlight regions of achromatic images, producing a perceptual grouping. In addiction, we suggest that those regions, 'survival' through different scales, coincide with the most prominent regions in the image.","PeriodicalId":39889,"journal":{"name":"Psicoterapia Cognitiva e Comportamentale","volume":"46 1","pages":"75-96"},"PeriodicalIF":0.0000,"publicationDate":"2001-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Psicoterapia Cognitiva e Comportamentale","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1174/021435501753635596","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"Psychology","Score":null,"Total":0}
引用次数: 0
Abstract
espanolLa nocion de mapa, entendida como correspondencia biunivoca entre dos niveles de representacion en el Sistema Visual (SV), nos lleva a plantear la cuestion de como puede utilizar la vision estos mapas (que incluyen las representaciones distribuidas) para procesar la informacion de la imagen. Para ello, fijamos dos objetivos. Por un lado, analizar la retina cortical, un mapa ubicado en el cortex estriado, que sirve de soporte a una importante representacion conjunta espacio-espectral. Para generar esta representacion concebimos un modelo, simplificado, compuesto por cuatro laminas de frecuencia espacial y cuatro canales de orientacion, el cual produce dicha representacion mediante Analisis espectral local. Por otro lado, proponemos otro modelo que permite la deteccion automatica de regiones de interes visual. Este modelo se fundamenta en la hipotesis de que las respuestas a los filtros lineales locales aplicados a la imagen, y almacenadas en la retina cortical, enalan las regiones que mas destacan en la imagen acromatica, produciendo, asi, un grupo perceptivo. Complementariamente, sugerimos que aquellas regiones que `perduran' en las diferentes escalas coinciden con las de mayor saliencia visual EnglishWe understand the term map as one-to-one correspondence between two representation levels in the Visual System (VS). This notion lead us to outline the question about how the (Visual System) VS can use such map (which include distributed representations) for the image processing. In order to achieve this goal we fix two objectives. On the one, to analyse the cortical retinae, a map located in the striate cortex that contains an important joint space/spectral representation, we have elaborated a simplified model to obtain this joint representation, which is composed by layers of cells tuned to different spatial frequencies and for columns tuned to different orienttions. The joint representation is generated by applying Local Spectral Analysis to an image. On the other, we proposed a new model that allow us to detect, automatically, interesting regions in the image. It is based in the following hypothesis: the outputs of linear and local filters applied to the image, and stored in the cortical retinae, indicate the more highlight regions of achromatic images, producing a perceptual grouping. In addiction, we suggest that those regions, 'survival' through different scales, coincide with the most prominent regions in the image.
西班牙地图概念,被理解为视觉系统(SV)中两个表示层次之间的双向对应,使我们提出了视觉如何使用这些地图(包括分布式表示)来处理图像信息的问题。为此,我们设定了两个目标。首先,分析皮质视网膜,这是一幅位于纹状皮层的地图,它支持一个重要的空间-光谱联合表示。为了生成这种表示,我们设计了一个由四个空间频率板和四个定向通道组成的简化模型,通过局部光谱分析生成这种表示。另一方面,我们提出了一种自动检测视觉感兴趣区域的模型。该模型基于一个假设,即应用于图像并存储在皮质视网膜中的局部线性滤波器的反应,激活消色差图像中最突出的区域,从而产生一个感知群体。此外,我们认为那些在不同尺度上“持久”的区域与那些视觉显著性最高的区域相匹配。我们将术语图理解为视觉系统(VS)中两个表示层次之间的一对一对应。这一概念使我们概述了(视觉系统)VS如何使用这种地图(包括分布式表示)进行图像处理的问题。为了实现这一目标,我们确定了两个目标。皮层On the one to analyse the map retinae, located in the striate皮质that contains joint space / spectral representation, we have an important阐明simplified model to obtain this joint representation, which is任by layers of / tuned to不同空间不同orienttions frequencies and for columns tuned to。= =地理= =根据美国人口普查,这个县的总面积是,其中土地和(3.064平方公里)水。另一方面,我们提出了一个新的模型,使我们能够自动检测图像中感兴趣的区域。It is based in the hypothesis):线性的研究所体制和地方filters applied to the image,皮层和存储retinae,表明the more强调区域of achromatic images,的perceptual grouping。在addiction中,我们认为这些通过不同尺度“生存”的区域与图像中最突出的区域相匹配。
期刊介绍:
Psicoterapia Cognitiva e Comportamentale, rivista quadrimestrale, pubblica articoli originali che contribuiscono allo sviluppo delle conoscenze teoriche ed al progresso della prassi clinica in psicoterapia cognitiva comportamentale. Ricadono nell’ambito d’interesse della rivista le applicazioni delle metodologie cognitive e comportamentali all’individuo, alla famiglia, al gruppo delle organizzazioni. Nell’ambito d’interesse sono parimenti incluse la valutazione clinica (assessment), la medicina comportamentale, la riabilitazione, la metodologia, la ricerca "analogica", la ricerca di base e la ricerca interdisciplinare rilevante per la psicoterapia.