Local Linear Transforms for Texture Measurements

نویسنده

  • Michael UNSER
چکیده

The Nth order probability density function for pixels in a restricted neighborhood may be characterized by a set of N histograms (or some corresponding moments) computed along appropriately chosen axes. The projections on those axes are obtained from a local linear transform of the local neighborhood vector. This approach is closely related to filter bank analysis methods and gives a statistical justification for the extraction of texture properties by means of convolution operators or local matches. Optimal and sub-optimal linear operators are derived for texture analysis and classification. Experimental results indicate that the method is robust, flexible, and that it performs as well as standard co-occurrence based methods for texture classification. The proposed approach enables texture characterization with a lower number of features and it is also computationally more appealing. Zusammeafassung. Das Verfahren der Bildanalyse mit Hilfe lokaler linearer Transformationen erlaubt es, die N-dimensionale Verteilungsdichtefunktion der Punkte eines begrenzten Bildausschnitts durch N Histogramme anzun~ihern, die entlang geeignet gew~ihlter Achsen aufgestellt werden. Die Projektionen auf diese Achsen werden mit Hilfe einer linearen Transformation des sogenannten Nachbarschaftsvektors berechnet. Diese Ann~iherung entspricht der Analyse mit Hilfe einer Filterbank und gibt eine statistische Rechtfertigung fiir die Extraktion yon Eigenschaften der Bildtextur mit Hilfe lokaler Merkmalsfilter. Optimale und suboptimale Lrsungen fiir die Wahl derartiger linearer Filter werden vorgeschlagen fiir Texturanalyse and Klassifikation. Wie Experimente mit realen Bildtexturen zeigen, ist die Methode unempfindlich, anpassbar, und ebenso zuverl/issig wie Standardmethoden die Pixelpaarebeziehungen fiir Texturklassifikation verwenden. Die vorgeschlagene Methode ermrglicht Texturbeschreibungen mit weniger Parametern und benrtigt weniger Computeroperationen. R~umr. La mrthode d'analyse de texture par transformation linfaire locale permet une caractfrisation partielle d 'une densit6 de probabilit6 d'ordre N par N histogrammes calculrs selon des axes convenablement choisis. Cette approche est 6quivalente une analyse par banc de filtres et apporte une justification statistique quant au principe de l 'extraction de proprirtrs de texture par des masques de convolution. Des solutions optimales et sous-optimales pour le choix des filtres sont proposres pour l'analyse et la classification de textures. On montre exprrimentalement que la mrthode est robuste et flexible. Pour la classification de textures, elle permet d'atteindre des performances aussi bonnes que les m&hodes usuelles se basant sur des mesures de co-occurrences. De plus, elle donne lieu/l une caractrrisation des proprirtrs de texture avec un nombre moindre d'attributs; elle se prate 6galement h u n calcul plus aisr.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Region Completion in a Texture using Multiresolution Transforms

Abstract Natural images, textures and photographs are likely to be impaired by stains.  As a result a substantial portion of the image remains blurred. However, a method called region completion is adopted to fill in the tainted part by using the information from the portion left unblemished by stains. A novel method to perform this operation is proposed in this paper. The three significant sta...

متن کامل

On texture analysis: Local energy transforms versus quadrature filters

The well-known method proposed by Laws for texture analysis first subjects the texture to a filter bank, followed by the computation of energy measures, e.g. through local variance estimation. As shown by Unser in 1986, the f i l ter bank application is equivalent to a linear transformation of the grey values of neighbouring pixels. In this contribution, we derive a further linear relationship ...

متن کامل

Performance Comparison of Gradient Mask Texture based Image Retrieval Techniques using Global and Local Hybrid Wavelet Transforms with Ternary Image Maps

The theme of the work presented here is performance comparison of gradient mask texture based image retrieval techniques using global and local hybrid wavelet transforms generated from the combination of Walsh, Haar and Kekre transforms. Ternary image maps of Prewitt/Robert/Sobel filtered images are compared with '64-pattern' texture set generated using local and global hybrid wavelet...

متن کامل

Gabor-like Eigenfilters for Texture Segmentation

Classification, segmentation and discrimination of textures are some of the most fundamental tasks in many image understanding, pattern recognition and machine vision applications. In the past two decades or so, these topics have been actively studied by many researchers and different approaches to the problem have been proposed. Earlier approaches to texture segmentation include co-occurrence ...

متن کامل

A Comparative Study of Filter Based Texture Operators Using Mahalanobis Distance

Texture feature extraction operators, which comprise linear filtering, eventually followed by post-processing, are considered. The filters used are Laws’ masks, filters derived from well-known discrete transforms, and Gabor filters. The post-processing step comprises non-linear point operations and/or local statistics computation. The performance is measured by means of the Mahalanobis distance...

متن کامل

New Face Recognition Method Based on Texture Features using Linear Wavelet Transforms

The present paper proposes a novel scheme of face recognition system based on texture features derived from co-occurrence parameters using 1-level linear wavelet decomposition technique. The present method divides the face into four parts. The novelty of the present scheme lies on usage of distance function on texture features for the recognition of facial images. The distance function evaluate...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2003