International Journal of Image Mining (5 papers in press)
Comparative Study of Different Machine Learning Classifiers for Mammograms and Brain MRI Images
by Poonam Sonar, Udhav Bhosle, Chandrajit Choudhury
Abstract: Abstract: Today, Breast cancer in women has become a leading cause of cancer deaths. Till date, Mammography is the most reliable and accurate technique for early and accurate detection of breast cancer. Therefore, the researchers are giving highest priority to computer aided diagnosis of breast cancer through mammograms. This paper presents the machine learning based mammogram classification techniques. Mammogram database images are pre-processed to extract region of interest. GLCM (grey level covariance matrix) based texture features are extracted from segmented ROI. These features are used to trained classifier. The trained classifier is used to classify breast tissues in normal / abnormal class and further to benign / malignant class. Different machine learning classifiers such as SVM, KNN, Random Forest 4.5, Logistic Regression, Fisher Discriminant analysis, Na
Keywords: Keywords: Mammograms; Texture feature ; SVM; KNN; Hybrid SVM_KNN; Random Forest.
Hardware Implementation of Stereo Vision Algorithms for Depth Estimation
by Nitish Wadne
Abstract: Depth estimation has applications like robot navigation, advance driver assistance systems, 3D movies etc. Depth is represented in terms of disparity map which can be generated using various stereo correspondence algorithms. This paper presents an implementation of semi global block matching algorithm on raspberry pi to estimate the depth from the camera. The algorithm computes the disparity using block wise matching and smoothness constraint. The proposed algorithm is compared with SAD algorithm on personal computer as well as on raspberry pi. The algorithm is also further, evaluated on the standard dataset. The project aim is to detect people in an image and estimate their depth from the camera. The real time implementation of the proposed algorithm uses block size of 21 x21 for images which has resolution of 1280 x 720 P. The algorithm estimates depth with an accuracy of 95%. The system also provides faster processing time to the proposed algorithm.
Keywords: Computer vision; depth estimation; Raspberry Pi; Semi-global block matching.
Vehicle Detection in Wide-Area Aerial Imagery: Cross-Association of Detection Schemes with Post-Processings
by Xin Gao
Abstract: Post-processing schemes are crucial for object detection algorithms to improve the performance of detection in wide-area aerial imagery. We select appropriate parameters for three algorithms (variational minimax optimization , feature density estimation  and Zhengs scheme by morphological filtering ) to achieve the highest average F-score on random sample frames, and then follow the same procedure to implement five post-processing schemes on each algorithm. Two low-resolution aerial videos are used as our datasets to compare automatic detection results with the ground truth objects on each frame. The performance analysis of post-processing schemes on each algorithm are presented under two sets of evaluation metrics.
Keywords: Post-processing; object detection; wide-area aerial imagery.
A Promising Method for Early Detection of Ischemic Stroke Area on Brain CT Images
by Amina Fatima Zahra Yahiaoui, Abdelhafid Bessaid
Abstract: Non-Contrast Computed Tomography (NCCT) has been chosen as the modality of choice for stroke imaging due to its low price and high availability. However, subtle changes of ischemia are hard to visualize and to extract on brain CT images. Alberta Stroke Program Early CT Score (ASPECTS) has been developed to help radiologists to make decisions regarding thrombolytic treatment. Only patients with favorable baseline scans (ASPECTS, 810) benefitted from endovascular revascularization therapy. The purpose of this study was to develop a novel approach for automated detection of ischemic stroke area on brain CT images within earliest hours after onset symptoms using comparison of brain hemispheres. The algorithm of the proposed method has five steps: preprocessing, segmentation of 10 Regions of Interest (ROIs), elimination of old infarcts and cerebrospinal fluid (CSF) space, feature extraction and stroke detection and ASPECTS scoring. The features obtained from ten ROIs were then used to select the abnormal regions and then to compute the corresponding ASPECTS score. The method was applied to 25 patients with infarctions of Middle Cerebral Artery (MCA) who presented to LA MEKERRA imaging center. The proposed method gives an effective results comparing with an existing method and a high sensitivity 90.8%. Our approach has the potential to be used as second opinion in stroke diagnosis.rnrn
Keywords: Stroke detection; Computed Tomography; ASPECTS score; bilateral comparison.
Special Issue on: Medical Imaging
DECISION BASED FUZZY LOGIC APPROACH FOR MULTIMODAL MEDICAL IMAGE FUSION IN NSCT DOMAIN
by Sivasangumani Selvaraj
Abstract: Image fusion is used to reduce the redundancy and increases the needed information in the processed image from two or more input images that have different information generated by different sources. The output image has more information and is more suitable for visual perception or processing tasks like medical imaging, remote sensing, concealed weapon detection, weather forecasting, biometrics etc. Image fusion methods basically accept only registered images to produce a high quality fused single image with spatial and spectral information. The fused image with more information will improve the performance of image analysis algorithms used in medical applications. In this paper, we proposed an image fusion algorithm based on decision approach and NSCT to improve the future resolution of the images. In this, images will be segmented into regions and decomposed into sub-images and then processed using Fuzzy Logic, the information fusion is performed using these images under the certain criteria such as non subsampled contourlet transform (NSCT) and certain fusion rules such as Fuzzy Logic, and finally these sub-images are reconstructed into the resultant image with plentiful information. The various metrices entropy, mutual information (MI) and Fusion Quality are calculated to compare the results. The proposed method is compared both subjectively as well as objectively with the other image fusion methods. The experimental results show that the proposed method is better than other fusion methods and increases the quality and PSNR of fused image.
Keywords: image fusion; discrete wavelet transform; DWT; NSCT; non subsampled contourlet transform; fuzzy logic; genetics algorithm; entropy; mutual information; MI.