Feature extraction is an important step in Computer Assisted Diagnosis of brain abnormalities using Magnetic Resonance Images (MRI).Feature Extraction is the process of reducing the size of image data by obtaining necessary information from the segmented image. The visual content of a segmented image can be captured using this process. From the extracted features it is possible to demarcate between normal and abnormal brain MRI. The reliability of the classification algorithm depends on segmentation method and extracted features. In this work texture features are extracted using Gray Level Co-occurrence Matrix (GLCM) and shape features are extracted using connected regions. Images with malignant tumor, benign tumor and normal brain have different features. This variation in feature values is useful in classification of MR images. The features thus obtained will be given to a classifier for training and testing. Keywords-Brain tumor, Magnetic resonance imaging (MRI), Computer aided diagnosis (CAD), Feature Extraction, Gray Level Co-occurrence Matrix (GLCM), Connected Regions. I. INTRODUCTION Images obtained from an MRI Scanner are verified by a radiologist for the identification of abnormalities. Manual classification of MR images is a time consuming and challenging task [8]. Hence there is a necessity for a computer aided automatic tumor classification system. Major step involved in computer aided detection of brain abnormalities are pre-processing, segmentation, feature extraction and classification.After segmenting the tumor regions from the MRI, the features of the segmented regions are analysed. The properties which provide description about the whole image are called features [3]. Feature extraction process reduces the original MRI data set into a set of features. This feature set is also known a feature vector. These feature vectors are the basic inputs for any classification algorithm. Based on the severity of tumors, they are classified into Benign and Malignant [7]. Benign tumors are slow growing and less harmful tumors. Malignant tumors grow fast and affect surrounding tissues [7]. Particular features exhibited by these tumors are useful in the classification process. Rajesh et al [9] used rough set theory for the extraction of features. The features thus extracted provided a classification efficiency of 90%.Hiremath et al [13] introduced a feature extraction technique based on complementary wavelet transform. In this method features are extracted from 4 sub bands and the efficiency is more compared to features from a single band. Discrete Wavelet Transform (DWT) was used and the accuracy of classification is less compared to other methods. Huang et al [14] introduced a method for dimensionality reduction using sub band grouping and selection. Low classification accuracy and less efficiency are the demerits of this method.Ramteke et al [10] obtained statistical texture features from input dataset. The classification efficiency obtained was 80%. Xuan et al [12] used texture symmetry and intensity based f...
Image Registration (IR) is the process of transformation of different data into the coordinate system and provides the geometric alignment of two images used in the computer vision, medical imaging and remote sensing applications. An image registration is an important stage in multi-temporal image processing since, the recovery of information from cloud shadow is difficult. Traditionally, the Demons, Combined Registration and Segmentation (CRS) approach, Markov Random Field (MRF) and Mutual Information (MI) based approaches offers more computational complexity, minimum edge preservation measure (QAB/F) during image registration process. To maximize the quality of edge preservation measure and MI with minimum computational time, this paper proposes hybrid Particle Swarm Optimization (PSO)-Affine Transformation (AT) technique for an image registration. An enhanced registration process and the cloud removal technique are proposed for quality improvement of an image. Initially, Gaussian filtering in the preprocessing stage removes the noises present in an image. The proposed PSO extracts the matching points between the reference image and target image in the multitemporal image dataset. Then, the AT on extracted matching points provides the specific feature points from main features. Finally, the Relevance Vector Machine (RVM) classification forms the cluster of specific feature points. The extracted feature points from PSO-AT maximize the quality of edge preservation and MI with efficient cloud removal. The comparative analysis with the traditional methods of Control Point -Least Square (CP-LS), MultiFocus Image Fusion (MFIF) and Discrete Wavelet Transform (DWT) on the parameters of QAB/F and MI shows the effectiveness of proposed PSO-AT.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.