Mausumi Acharyya

Document image segmentation using wavelet scale-space features

An efficient and computationally fast method for segmenting text and graphics part of document images based on textural cues is presented. We assume that the graphics part have different textural properties than the nongraphics (text) part. The segmentation method uses the notion of multiscale wavelet analysis and statistical pattern recognition. We have used M-band wavelets which decompose an image into M/spl times/M bandpass channels. Various combinations of these channels represent the image at different scales and orientations in the frequency plane. The objective is to transform the edges between textures into detectable discontinuities and create the feature maps which give a measure of the local energy around each pixel at different scales. From these feature maps, a scale-space signature is derived, which is the vector of features at different scales taken at each single pixel in an image. We achieve segmentation by simple analysis of the scale-space signature with traditional k- means clustering. We do not assume any a priori information regarding the font size, scanning resolution, type of layout, etc. of the document in our segmentation scheme.

Segmentation of remotely sensed images using wavelet features and their evaluation in soft computing framework

The present paper describes a feature extraction method based on M-band wavelet packet frames for segmenting remotely sensed images. These wavelet features are then evaluated and selected using an efficient neurofuzzy algorithm. Both the feature extraction and neurofuzzy feature evaluation methods are unsupervised, and they do not require the knowledge of the number and distribution of classes corresponding to various land covers in remotely sensed images. The effectiveness of the methodology is demonstrated on two four-band Indian Remote Sensing 1A satellite (IRS-1A) images containing five to six overlapping classes and a three-band SPOT image containing seven overlapping classes.

Extraction of features using M-band wavelet packet frame and their neuro-fuzzy evaluation for multitexture segmentation

In this paper, we propose a scheme for segmentation of multitexture images. The methodology involves extraction of texture features using an overcomplete wavelet decomposition scheme called discrete M-band wavelet packet frame (DMbWPF). This is followed by the selection of important features using a neuro-fuzzy algorithm under unsupervised learning. A computationally efficient search procedure is developed for finding the optimal basis based on some maximum criterion of textural measures derived from the statistical parameters for each of the subbands. The superior discriminating capability of the extracted features for segmentation of various texture images over those obtained by several existing methods is established.

M-BAND WAVELETS: APPLICATION TO TEXTURE SEGMENTATION FOR REAL LIFE IMAGE ANALYSIS

This paper describes two examples of real-life applications of texture segmentation using M-band wavelets. In the rst part of the paper, an ecien t and computationally fast method for segmenting text and graphics part of a document image based on textural cues is presented. It is logical to assume that the graphics part has dieren t textural properties than the non-graphics (text) part. So, this is basically a two-class texture segmentation problem. The second part of the paper describes a segmentation scheme for another real-life data such as remotely sensed image. Dieren t quasi-homogeneous regions in the image can be treated to have dieren t texture properties. Based on this assumption the multi-class texture segmentation scheme is applied for this purpose.

EXTRACTION OF NOISE TOLERANT, GRAY-SCALE TRANSFORM AND ROTATION INVARIANT FEATURES FOR TEXTURE SEGMENTATION USING WAVELET FRAMES

In this paper, we propose a texture feature extraction scheme at multiple scales and discuss the issues of rotation and gray-scale transform invariance as well as noise tolerance of a texture analysis system. The nonseparable discrete wavelet frame analysis is employed which gives an overcomplete wavelet decomposition of the image. The texture is decomposed into a set of frequency channels by a circularly symmetric wavelet filter, which in essence gives a measure of edge magnitudes of the texture at different scales. The texture is characterized by local energies over small overlapping windows around each pixel at different scales. The features so extracted are used for the purpose of multi-texture segmentation. A simple clustering algorithm is applied to this signature to achieve the desired segmentation. The performance of the segmentation algorithm is evaluated through extensive testing over various types of test images.

Adaptive basis selection for multi texture segmentation by M-band wavelet packet frames

We propose an approach for texture feature extraction based on M-band wavelet packet frames. The features so extracted are used for segmentation of multi texture images. Standard dyadic wavelets are not suitable for the analysis of high frequency signals with relatively narrow bandwidth and also are not translation invariant. Also, since most significant information of a texture often lies in the intermediate frequency bands, the present work employs an overcomplete wavelet decomposition scheme called discrete M-band wavelet packet frame (DM-bWPF), which yields improved segmentation accuracies. Wavelet packets represent a generalization of the method of multiresolution decomposition and comprise all possible combinations of subband tree decomposition. We propose a computationally efficient search procedure to find the optimal basis based on some maximum criterion of textural measures derived from the statistical parameters of each of the subbands, to locate dominant information in each subband (frequency channel) and decide further decomposition.

Wavelet-based texture segmentation of remotely sensed images

A texture feature extraction scheme based on M-band wavelet packet frames is investigated. The features so extracted are used for segmentation of satellite images which usually have complex and overlapping boundaries. The underlying principle is based on the fact that different image regions exhibit different textures. Since most significant information of a texture often lies in the intermediate frequency bands, the present work employs an overcomplete wavelet decomposition scheme called discrete M-band wavelet packet frame (DM-bWPF), which yields improved segmentation accuracies. Wavelet packets represent a generalization of the method of multiresolution decomposition and comprise all possible combinations of subband tree decomposition. We propose a computationally efficient search procedure to find the optimal basis based on some maximum criterion of textural measures derived from the statistical parameters of each of the subbands, to locate dominant information in each subbands (frequency channels) and decide further decomposition.

Multiscale Segmentation of Document Images Using M -Band Wavelets

In this work we propose an algorithm for segmentation of the text and non-text parts of document image using multiscale feature vectors. We assume that the text and non-text parts have different textural properties. M-band wavelets are used as the feature extractors and the features give measures of local energies at different scales and orientations around each pixel of the M×M bandpass channel outputs. The resulting multiscale feature vectors are classified by an unsupervised clustering algorithm to achieve the required segmentation, assuming no a priori information regarding the font size, scanning resolution, type layout etc. of the document.

Two texture segmentation using M-band wavelet transform

The M-band wavelet decomposition, which is a direct generalization of the standard 2-band wavelet decomposition has been applied to the problem of an unsupervised segmentation of two texture systems. Standard wavelets are not suitable for the analysis of high frequency signals with relatively narrow bandwidth. So we propose to use the decomposition scheme based on M-band wavelets, that yield improved segmentation accuracies. Unlike the standard wavelet decomposition which gives a logarithmic frequency resolution, the M-band decomposition gives a mixture of a logarithmic and linear frequency resolution. Further motivation to use the M-band wavelet filter for texture analysis is because this decomposition yields a large number of sub-bands which is required for good quality segmentation.

Classification of Rotated Textures using Overcomplete Wavelet Frames

In this paper we present an approach to characterize textures at multiple scales using wavelet transforms and propose a texture classification algorithm that is invariant to rotation and translation. The nonseparable discrete wavelet frame is used as the wavelet transform that decompose the texture images into a set of frequency channels. In each channel we take the variance as feature. Classification experiments using twenty Brodazt textures indicate that texture signatures based on wavelet frame analysis are beneficial for accomplishing subtle discrimination of textures and robust classification against rotation and translation.