Fátima N. S. de Medeiros

Edge preserving wavelet speckle filtering

A new edge preserving wavelet filtering is proposed and it is applied to real SAR images that are generally affected by a multiplicative noise, called speckle, which degrades the quality of these images. The new approach attempts to look for the neighborhood area in the detail images of a wavelet decomposition, that identifies homogeneous areas and edge information by using masks in order to reduce speckle while edges are preserved The improved filtering method uses the Nagao and Matsuyama and Tomita and Tsuji masks to detect edge locations in wavelet subspaces. The information provided by the masks is used to distinguish which of the detail coefficients, are to be shrunk.

An unsupervised coarse-to-fine algorithm for blood vessel segmentation in fundus images

An unsupervised method to detect blood vessels in fundus images is proposed.The algorithm effectively tackles image distortions such as central vessel reflex.The two expert vessel identification images present significant differences.The average observer plays an important role in defining a neutral standard.Balanced accuracy is an alternative for performance evaluation of segmentation. Algorithms for retinal vessel segmentation are powerful tools in automatic tracking systems for early detection of ophthalmological and cardiovascular diseases, and for biometric identification. In order to create more robust and reliable systems, the algorithms need to be accurately evaluated to certify their ability to emulate specific human expertise. The main contribution of this paper is an unsupervised method to detect blood vessels in fundus images using a coarse-to-fine approach. Our methodology combines Gaussian smoothing, a morphological top-hat operator, and vessel contrast enhancement for background homogenization and noise reduction. Here, statistics of spatial dependency and probability are used to coarsely approximate the vessel map with an adaptive local thresholding scheme. The coarse segmentation is then refined through curvature analysis and morphological reconstruction to reduce pixel mislabeling and better estimate the retinal vessel tree. The method was evaluated in terms of its sensitivity, specificity and balanced accuracy. Extensive experiments have been conducted on DRIVE and STARE public retinal images databases. Comparisons with state-of-the-art methods revealed that our method outperformed most recent methods in terms of sensitivity and balanced accuracy with an average of 0.7819 and 0.8702, respectively. Also, the proposed method outperformed state-of-the-art methods when evaluating only pathological images that is a more challenging task. The method achieved for this set of images an average of 0.7842 and 0.8662 for sensitivity and balanced accuracy, respectively. Visual inspection also revealed that the proposed approach effectively addressed main image distortions by reducing mislabeling of central vessel reflex regions and false-positive detection of pathological patterns. These improvements indicate the ability of the method to accurately approximate the vessel tree with reduced visual interference of pathological patterns and vessel-like structures. Therefore, our method has the potential for supporting expert systems in screening, diagnosis and treatment of ophthalmological diseases, and furthermore for personal recognition based on retinal profile matching.

Locating oil spill in SAR images using wavelets and region growing

One of the difficulties of using Artificial Neural Networks (ANNs) to estimate atmospheric temperature is the large number of potential input variables available. In this study, four different feature extraction methods were used to reduce the input vector to train four networks to estimate temperature at different atmospheric levels. The four techniques used were: genetic algorithms (GA), coefficient of determination (CoD), mutual information (MI) and simple neural analysis (SNA). The results demonstrate that of the four methods used for this data set, mutual information and simple neural analysis can generate networks that have a smaller input parameter set, while still maintaining a high degree of accuracy.

Filtering effects on SAR images segmentation

This paper evaluates filtering effects on SAR image segmentation testing four types of speckle reduction algorithms. A primary goal of these filters is to provide a large amount of speckle noise reduction in homogeneous areas and to preserve edges and details. To assess the effects produced by the filters in the posterior segmentation task, some quality measures are calculated from the processed images and used to indicate the filtering ability for features preservation.

Speckle noise MAP filtering based on local adaptive neighborhood statistics

This work proposes the use of an adaptive neighborhood procedure to extract local statistical properties of images in order to improve a speckle noise Maximum a Posteriori (MAP) filtering performance. The strategy consists in growing the local statistically homogeneous area near the pixel in order to estimate its MAP filtering parameters. Measures evaluating both the signal-to-noise improvement and resolution loss due to filtering are computed. The use of region growing is investigated as a promising approach compared to a fixed size and shape neighborhood.

Combined use of MAP estimation and K-means classifier for speckle noise filtering in SAR images

The main purpose of this work is to study and implement a maximum a posteriori (MAP) filter combined with the K-means algorithm in order to reduce speckle noise in SAR images. The K-means algorithm over Lis (1988) coefficient is used to classify the noisy image in regions of homogenous statistics. This kind of information is used as a guide for choosing the best window size for parameter estimation in the MAP filtering. This paper is based on the multiplicative model for speckle and considers different densities to describe the a priori knowledge. It suggests a new adaptive filtering algorithm based on the MAP approach and clustering.

Adaptive speckle MAP filtering for SAR images using statistical clustering

This paper presents a nonlinear adaptive filter based on the the maximum a posteriori (MAP) approach to reduce speckle in one-look, linear detected SAR images. The k-means clustering algorithm is combined with the MAP filter in order to cluster pixels with similar statistics (Changle Lis variance ratio). Assigned to each cluster there is a window size which is used to estimate the filter parameters. Several densities such as gaussian, gamma, chi-square, exponential, and Rayleigh were used as a priori model. To assess the improvement brought by the proposed algorithm we evaluate it with respect to edge preservation via Hough transform.

Evaluating an adaptive windowing scheme in speckle noise MAP filtering

Synthetic aperture radar (SAR) images are corrupted by speckle noise, which degrades the quality and interpretation of the images. Speckle removal provides a better interpretability of SAR images if the technique performs the filtering without loss of spatial resolution and preserves fine details and edges. This work aims to redefine the neighborhood areas around the noisy pixel and in this area the local mean and variance are computed to estimate the Maximum a Posteriori (MAP) filter parameters. The proposed modified MAP algorithm improves the ability to filter the speckle noise without blurring edges and targets by applying the MAP estimator in the current adaptive window that is controlled by a measure of homogeneity in the area around the noisy pixel. This adaptive windowing was also incorporated to the classical Kuan et al. (1985) and Frost et al. (1982) filters in order to evaluate the performance of the proposed scheme. The effectiveness in reducing speckle by the modified MAP filter is evaluated in terms of qualitative and quantitative aspects such as line and edge preservation and the improvement of the signal to noise ratio. The tests were performed in real SAR images.

Multiscale denoising algorithm based on the a trous algorithm

In this work we present a novel application of the multiscale denoising algorithm proposed by Sita and Ramakrishnan (2000). We used it to filter artificially contaminated images by multiplicative speckle and additive Gaussian noise, respectively. This filtering scheme is a combination of the shift invariant discrete wavelet and nonlinear filtering applied to evoked potential signals. It employs a redundant discrete wavelet (the a trous algorithm) removing the smallest wavelet coefficients in each dyadic scale guided by the correlation existing between them in different scales.

Automatic Detection of Fovea in Retinal Images Using Fusion of Color Bands

This paper presents a new method for fovea detection in color retinal images. Automatic detection of this anatomical structure is a prerequisite for computer aided diagnosis of several retinal diseases, such as macular degeneration. The proposed algorithm detects the macula center by determining a region of interest (ROI) and taking into account optic disk (OD) coordinates and the fact that the central region, i.e. fovea, is a homogenous dark area without blood vessels. Our segmentation algorithm searches for the lowest mean color intensity window in the enhanced image that results from a fusion between the red and green channels. Then, tests were carried on three public benchmark databases, which constitute a total of 254 images.