Banshidhar Majhi

Mammogram classification using two dimensional discrete wavelet transform and gray-level co-occurrence matrix for detection of breast cancer

In this paper, we propose a mammogram classification scheme to classify the breast tissues as normal, benign or malignant. Feature matrix is generated using GLCM to all the detailed coefficients from 2D-DWT of the region of interest (ROI) of a mammogram. To derive the relevant features from the feature matrix, we take the help of t-test and F-test separately. The relevant features are used in a BPNN classifier for classification. Two standard databases MIAS and DDSM are used for the validation of the proposed scheme. It is observed that t-test based relevant features outperforms to that of F-test with respect to accuracy. In addition to the suggested scheme, the competent schemes are also simulated for comparative analysis. It is observed that the proposed scheme has a better say with respect to accuracy and area under curve (AUC) of receiver operating characteristic (ROC). The accuracy measures are computed with respect to normal vs. abnormal and benign vs. malignant. For MIAS database these accuracy measures are 98.0% and 94.2% respectively, whereas for DDSM database they are 98.8% and 97.4%.

Robust iris indexing scheme using geometric hashing of SIFT keypoints

This paper proposes an efficient indexing scheme for searching large iris biometric database that achieves invariance to similarity transformations, illumination and occlusion. The proposed scheme considers local descriptors as well as relative spatial configuration for claiming identity. To overcome the effect of non-uniform illumination and partial occlusion due to eyelids, local features are extracted from noise independent annular iris image using scale invariant feature transform (SIFT). The detected keypoints are used to index iris database by applying geometric hashing scheme that is robust to similarity transformations as well as occlusion. During iris retrieval, geometric hashed location from query iris image is obtained to access the appropriate bin of hash table and for every entry found there, a vote is casted. The iris images that receive more than certain number of votes are considered as possible candidates. In order to find the potential matches, the keypoint descriptor of the list of possible candidates is matched with the query iris. Since only small portion of database is scanned to find a match it reduces the query retrieval time and improves accuracy. This approach is tested on UBIRIS, BATH, CASIA and IITK iris databases and shows a substantial improvement over exhaustive search technique in terms of time and accuracy.

Brain MR image classification using two-dimensional discrete wavelet transform and AdaBoost with random forests

This paper presents an automated and accurate computer-aided diagnosis (CAD) system for brain magnetic resonance (MR) image classification. The system first utilizes two-dimensional discrete wavelet transform (2D DWT) for extracting features from the images. After feature vector normalization, probabilistic principal component analysis (PPCA) is employed to reduce the dimensionality of the feature vector. The reduced features are applied to the classifier to categorize MR images into normal and abnormal. This scheme uses an AdaBoost algorithm with random forests as its base classifier. Three benchmark MR image datasets, Dataset-66, Dataset-160, and Dataset-255, have been used to validate the proposed system. A 5×5-fold stratified cross validation scheme is used to enhance the generalization capability of the proposed scheme. Simulation results are compared with the existing schemes and it is observed that the proposed scheme outperforms others in all the three datasets.

A new optimal delay and energy efficient coordination algorithm for WSAN

Wireless sensor and actor networks require efficient coordination mechanisms to perform reliable actions in the physical world. In this paper, an energy and delay aware three-level coordination mechanism is proposed for wireless sensor and actor networks using two-level hierarchical k-hop clustering algorithm. An optimal number of actors are calculated based on the network area and number of sensors. In the first level, sensors form a k-hop cluster by placing actor nodes as a cluster heads. In the second level, sink acts as the cluster head and forms a cluster among actors. A coordination mechanism among sensors and actors is also proposed based on their characteristics. To evaluate the performance of the proposed algorithm, it is simulated in NS2 and compared with existing algorithms. Simulation results reveal that the proposed algorithm outperforms the existing algorithms in various QoS parameters such as energy, delay and packet delivery ratio.

Improved offline signature verification scheme using feature point extraction method

In this paper a novel offline signature verification scheme has been proposed. The scheme is based on selecting 60 feature points from the geometric centre of the signature and compares them with the already trained feature points. The classification of the feature points utilizes statistical parameters like mean and variance. The suggested scheme discriminates between two types of originals and forged signatures. The method takes care of skill, simple and random forgeries. The objective of the work is to reduce the two vital parameters False Acceptance Rate (FAR) and False Rejection Rate (FRR) normally used in any signature verification scheme. In the end comparative analysis has been made with standard existing schemes.

Second order difference based detection and directional weighted median filter for removal of random valued impulsive noise

The proposed approach of removal of random valued impulsive noise from images works in two phases. The first phase detects contaminated pixels and the second phase filters only those pixels keeping others intact. The detection scheme utilizes second order difference of pixels in a test window and the filtering scheme is a variation median filter based on the edge information. The proposed scheme is simulated extensively on standard images and comparison with existing schemes reveal that our scheme outperforms them in terms of Peak Signal to Noise Ratio (PSNR), number of false detection and miss detection. The proposed scheme is also good at preserving finer details. Further, the computational complexity and number of iterations needed by the proposed scheme is less than the existing counterparts.

Fast segmentation and adaptive SURF descriptor for iris recognition

Abstract In this paper a robust segmentation and an adaptive SURF descriptor are proposed for iris recognition. Conventional recognition systems extract global features from the iris. However, global features are subject to change for transformation, occlusion and non-uniform illumination. The proposed iris recognition system handles these issues. The input iris image is used to remove specular highlights using an adaptive threshold. Further, the pupil and iris boundaries are localized using a spectrum image based approach. The annular region between the pupil and iris boundaries is transformed into an adaptive strip. The strip is enhanced using a gamma correction approach. Features are extracted from the adaptive strip using Speeded Up Robust Features (SURF). The results obtained using SURF are compared with the existing SIFT descriptor and the proposed approach performs with improved accuracy and reduced computation cost.

Indexing Iris Biometric Database Using Energy Histogram of DCT Subbands

The key concern of indexing is to retrieve small portion of database for searching the query. In the proposed paper iris database is indexed using energy histogram. The normalised iris image is divided into subbands using multiresolution DCT transformation. Energy based his- togram is formed for each subband using all the images in the database. Each histogram is divided into fixed size bins to group the iris images having similar energy value. The bin number for each subband is obtained and all subands are traversed in Morton order to form a global key for each image. During database preparation the key is used to traverse the B tree. The images with same key are stored in the same leaf node. For a given query image, the key is generated and tree is traversed to end up to a leaf node. The templates stored at the leaf node are retrieved and compared with the query template to find the best match. The proposed indexing scheme is showing considerably low penetration rate of 0.63%, 0.06% and 0.20% for CASIA, BATH and IITK iris databases respectively.

Intensity Range Based Background Subtraction for Effective Object Detection

In this letter, we propose an intensity range based object detection scheme for videos with fixed background and static cameras. The scheme suggests two different algorithms; the first one models the background from initial few frames and the second algorithm extracts the objects based on local thresholding. The strength of the scheme lies in its simplicity and the fact that, it defines an intensity range for each pixel location in the background to accommodate illumination variation as well as motion in the background. The efficacy of the scheme is shown through comparative analysis with competitive methods. Both visual as well as quantitative measures show an improved performance and the scheme has a strong potential for applications in real time surveillance.

Annular Iris Recognition Using SURF

This paper proposes an iris recognition system which can handle efficiently the problem of rotation, scaling, change in gaze of individual and partial occlusions that are inherent to non-restrictive iris imaging system. In addition to this, traditional iris normalisation approach deforms texture features linearly due to change in camera to eye distance or non-uniform illumination. To overcome the effect of aliasing features are extracted directly from annular region of iris using Speeded Up Robust Features (SURF). These features are invariant to transformations and occlusion. The system is tested on BATH, CASIA and IITK databases and is showing an accuracy of more than 97%. From the results it is inferred that local features from annular iris gives much better accuracy for poor quality images in comparison to normalised iris.