Akansha Mehrotra

A Fuzzy Kohonen Local Information C-Means Clustering for Remote Sensing Imagery

ABSTRACT This paper presents a neuro fuzzy clustering algorithm, Fuzzy Kohonen Local Information C-Means (FKLICM), for classification of remote sensing images. The proposed algorithm is a hybridization of the conventional Kohonen clustering network and Fuzzy Local Information C-Means (FLICM) to produce a much more efficient and accurate clustering algorithm. The proposed algorithm first forms a fused image with three Multispectral bands and pan band of Landsat 7 Enhanced Thematic Mapper Plus (ETM+) using the Brovey transform. The fused image is a three band image with higher resolution and better visual perception. The fused image is reduced to a one-dimensional image using principal component analysis (PCA). The FKLICM algorithm is applied on the PC-1 image to classify the remote sensing image into different land cover types. Integrating the neural network with a fuzzy system combines the advantages and overcomes the limitations of both technologies. The experimental results of the proposed algorithm are compared with two other algorithms, FCM and GIFP-FCM. The classification results and accuracy assessment show that FKLICM yields better results than the other methods.

A novel edge preserving filter for impulse noise removal

This paper proposes an edge preserving filter for removal of impulse noise. Digital images received from various sources are often degraded due to impulse noise and thus become unsuitable for further processing. To overcome this degradation removal of impulse noise is very important. In this paper an effective and efficient method of impulse noise removal is proposed which not only removes noise but also preserves edges. The algorithm first finds noisy, noise free and edge pixels. Then it replaces the noisy pixel with a pixel from its neighbourhood which is nearest to the adaptive median of the noisy pixel, this removes the noise as well as preserves edges and fine image details.

A Novel Algorithm for Impulse Noise Removal and Edge Detection

Edge detection of images is an important task in computer vision and image processing. Edge detection of noise free images is relatively simpler, but in most practical cases the images are degraded by noise. To find the edges from noisy images is a challenging task. This paper proposes a novel edge detection algorithm for images corrupted with noise. The algorithm finds the edges by eliminating the noise from the image so that the correct edges are determined. For making the image noise free the algorithm calculates closeness parameters, based on this parameter the noisy pixel is replaced by the most appropriate value. The edges of the noise free image are determined using morphological operators erosion and dilation. The proposed algorithm uses a combination of these operators to find the edges. This algorithm uses two different types of structuring elements so that all the edges of the image are determined efficiently.

Unsupervised change detection from remote sensing images using hybrid genetic FCM

This paper presents a new technique for unsupervised change detection in bitemporal remote sensing images using spectral change difference images and hybrid genetic FCM. The proposed method works in three steps. In the first step, three spectral change difference images:absolute value difference image, ratio image and log ratio image are computed. In the next step, a feature vector space is created using PCA. Finally, the change detection is obtained by dividing the feature vector space into two clusters using genetic FCM. The validity of the clusters is measured by DB index. The parts of image of Reno-Lake Tahoe area was used as data set for the performance evaluation of proposed algorithm. The results obtained were compared with EM based, MRF based and NSCT methods. The results verify that the proposed algorithm provides superior results than the other existing methods.

A N 8 (P) detail preserving adaptive filter for impulse noise removal

This paper proposes a N8(p) detail preserving adaptive filter for impulse noise removal. Impulse noise degrade the digital images due to which these images cannot be used for high level processing. Thus, image restoration becomes important. In this paper an effective and efficient method of impulse noise removal is proposed which not only removes noise but also preserves image details. The algorithm first classifies all the pixels as noise and noise free based on its N8(p) neighbours using averaging parameters introduced here and then replaces the noise pixels by the adaptive median of the pixel. The algorithm uses adaptive median as it provides better denoising and since the proposed algorithm performs prior classification of pixels as noise and noise free this preserves image details.

Detection of tsunami-induced changes using generalized improved fuzzy radial basis function neural network

Abstract The coastal areas of Japan were hard hit by a magnitude 9.0 earthquake on 11 March 2011. The earthquake triggered a disastrous tsunami over the area which led to massive destruction. In this paper, tsunami-induced changes in Soma, Watari, Natori and Iwanuma areas using Landsat 7 ETM+ and EO-1 ALI images are identified. The proposed method is based on image classification using radial basis function neural network and generalized improved fuzzy partition FCM algorithm. The pre- and post-tsunami images of the area are first classified using a radial basis function neural network. The pre- and post-tsunami images are classified into three classes including water, vegetation and urban and bare land class. The classified images are compared with other to obtain a set of four change classes. These change classes are labelled to obtain a classified change map. The change map reveals that large areas of vegetations and urban land are washed away by the tsunami in all the four cities, Soma, Watari, Natori and Iwanuma. The accuracy assessment of the method shows that the results obtained are quite satisfactory. The method has high overall accuracy and kappa coefficient value.

Change Detection from Satellite Images Using PNN

This paper presents a supervised change detection technique for satellite images using a probabilistic neural network (PNN). The proposed method works in two phases. In the first phase a difference image is computed. The most commonly used techniques for computing the difference image such as ratio images or log ratio images degrade the performance of the algorithm in the presence of speckle noise. To overcome the above mentioned limitations the difference image in this work is computed using normalized neighborhood ratio based method. In the next phase the PNN is used to detect efficiently any change between the two images. An estimator is used by the PNN to estimate the probability density function. The ratio of two conditional probability density functions, called the likelihood ratio is computed. Finally, the log likelihood ratio test is used to classify the pixels of the difference image into changed and unchanged classes to create a change map. The change map highlights the changes that have occurred between the two input images. The proposed method was compared quantatively as well as qualitatively with other existing state of the art methods. The results showed that the proposed method outperforms the other methods.