No abstract

Human face detection is concerned with finding the location and size of every human face in a given image. Face detection plays a very important role in human computer interaction field. It represents the first step in a fully automatic face re cognition, facial features detection, and expression recognition. There are many techniques used in face detection, each one has its advantages and disadvantages. The face detection system presented in this paper is a hybrid of known algorithms. First stag e of the proposed method is applying skin detection algorithm to specify all skin locations in the image. Second, extract face features like eyes, mouth and nose. At the last, a verification step is applied to ensure that the extracted features are facial features. In experiments on images having upright frontal faces with any background our system has achieved high detection rates and low false positives.

Millions of women are suffering from breast cancer, which can be treated effectively if it is detected early. Mammography is broadly recognized as an effective imaging modality for the early detection of breast cancer. Computer-aided diagnosis (CAD) systems are very helpful for radiologists in detecting and diagnosing abnormalities earlier and faster than traditional screening programs. An important step of a CAD system is feature extraction. This research gives a comprehensive study of the effects of different features to be used in a CAD system for the classification of masses. The features are extracted using local binary pattern (LBP), which is a texture descriptor, statistical measures, and multi-resolution frameworks. Statistical and LBP features are extracted from each region of interest (ROI), taken from mammogram images, after dividing it into N×N blocks. The multi-resolution features are based on discrete wavelet transform (DWT) and contourlet transform (CT). In multi-resolution analysis, ROIs are decomposed into low sub-band and high sub-bands at different resolution levels and the coefficients of the low sub-band at the last level are taken as features. Support vector machines (SVM) is used for classification. The evaluation is performed using Digital Database for Screening Mammography (DDSM) database. An accuracy of 98.43 is obtained using statistical or LBP features but when both these types of features are fused, the accuracy is increased to 98.63. The CT features achieved classification accuracy of 98.43 whereas the accuracy resulted from DWT features is 96.93. The statistical analysis and ROC curves show that methods based on LBP, statistical measures and CT performs equally well and they not only outperform DWT based method but also other existing methods.

Digital Mammograms are currently the most effective imaging modality for early detection of breast cancer but the number of false negatives and false positives is high. Mass is one type of breast lesion and the detection of masses is highly challenged problem. Almost all methods that have been proposed so far suffer from high number of false positives and false negatives. In this paper, a method for detecting true masses is presented, especially, for the reduction of false positives and false negatives. The key idea of the proposal is the use of Gabor filter banks for extracting the most representative and discriminative local spatial textural properties of masses that are present in mammograms at different orientations and scales. The system is evaluated on 512 (256 normal+256 true mass) regions of interests (ROIs) extracted from digital mammograms of DDSM database. We performed experiments with Gabor filter banks having different numbers of orientations and scales to find the best parameter setting. Using a powerful feature selection technique and support vector machines (SVM) with 10-fold cross validation, we report to achieve Az = 0.995±0.011, the area under ROC. Comparison with state-of-the-art techniques suggests that the proposed system outperforms similar methods, which are based on texture description, and the difference is statistically significant.

No abstract