Stan Ipson

An efficient Image Retrieval through DCT Histogram Quantization

This paper proposes a new simple method of Discrete Cosine Transform (DCT) feature extraction that is used to accelerate the speed and decrease the storage needed in the image retrieving process. Image features are accessed and extracted directly from JPEG compressed domain. This method extracts and constructs a feature vector of histogram quantization from partial DCT coefficient in order to count the number of coefficients that have the same DCT coefficient over all image blocks. The database image and query image is equally divided into a non overlapping 8X8 block pixel, each of which is associated with a feature vector of histogram quantization derived directly from discrete cosine transform DCT. Users can select any query as the main theme of the query image. The retrieved images are those from the database that bear close resemblance with the query image and the similarity is ranked according to the closest similar measures computed by the Euclidean distance. The experimental results are significant and promising and show that our approach can easily identify main objects while to some extent reducing the influence of background in the image and in this way improves the performance of image retrieval.

One Scan Connected Component Labeling Technique

This paper, presents a new component labeling algorithm which is based on scanning and labeling the objects in a single scan. The algorithm has the ability to test the four and eight connected branches of the object. This algorithm, which is fast and requires low memory allocation, can also process an image that contains large numbers of objects. The algorithm is used to scan the image from left to right and from top to bottom to find the unlabeled objects. A comparison analysis is performed with other component labeling algorithms. Our algorithm has shown an outstanding performance with respect to the processing time. A practical application with computer based mammography is also included.

Face detection based on skin color in image by neural networks

Face detection is one of the challenging problems in the image processing. A novel face detection system is presented in this paper. The approach relies on skin based color, while features extracted from two dimentional discreate cosine transfer (DCT) and neural networks. which can be used to detect faces by using skin color from DCT coefficient of Cb and Cr feature vectors. This system contains the skin color which is the main feature of faces for detection and then the skin face candidate is examined by using the neural networks, which learns from the feature of faces to classify whether the original image includes a face or not. The processing stage is based on normalization and discreate cosine transfer ( DCT ). Finally the classification based on neural networks approach. The experiments results on upright frontal color face images from the internet show an a excellent detection rate.

Face detection based neural networks using robust skin color segmentation

This paper proposes a robust schema for face detection system via Gaussian mixture model to segment image based on skin color. After skin and non skin face candidatespsila selection, features are extracted directly from discrete cosine transform (DCT) coefficients computed from these candidates. Moreover, the back-propagation neural networks are used to train and classify faces based on DCT feature coefficients in Cb and Cr color spaces. This schema utilizes the skin color information, which is the main feature of face detection. DCT feature values of faces, representing the data set of skin/non-skin face candidates obtained from Gaussian mixture model are fed into the back-propagation neural networks to classify whether the original image includes a face or not. Experimental results shows that the proposed schema is reliable for face detection, and pattern features are detected and classified accurately by the backpropagation neural networks.

A new approach for off-line handwritten Arabic word recognition using KNN classifier

Due to similarities between Arabic letters, and the various writing styles employed, recognition of Arabic handwritten text can be difficult. In this paper, an off-line Arabic handwritten word recognition system is proposed, in which technical details are presented in terms of three stages, i.e. preprocessing, feature extraction and classification. Firstly, words are segmented from input scripts and also normalized in size. Secondly, each segmented word is divided into overlapping blocks. Absolute mean values computed for each block of segmented words constitutes a feature vector. Finally, the resulting feature vectors are used to classify the words using the K nearest Neighbour classifier (KNN). The proposed system has been successfully tested on the IFN/ENIT database consisting of 32492 Arabic handwritten words which are written by more than 1000 different writers. Experimental results show a good recognition rate when compared with other methods.

Comparisons of feature selection methods using discrete wavelet transforms and Support Vector Machines for mammogram images

In this paper, we investigate wavelet-based feature extraction from mammogram images and efficient dimensionality reduction techniques. The aim is to propose a new computerized feature extraction technique to identify abnormalities in breast mammogram images. In this work, dimensionality reduction is carried out using the minimal-redundancy-maximal-relevance criterion (mRMR). The classification accuracy for each set of features is measured and evaluated using machine learning techniques and support vector machines (SVMs).

Texture-Based Feature Extraction for the Microcalcification from Digital Mammogram Images

This paper describes our ongoing efforts to provide efficient and accurate classification of microcalcification clusters in mammogram images. In this paper, a study of the characteristics of true microcalcifications compared to falsely detected microcalcifications is carried out using first and second order statistical texture analysis techniques. These features are generated in order to reduce the false positive (FP) ratio for the mammogram images. The statistical method presented here can successfully reduce the ratio of false positives (FP) by 18% without affecting the ratio of true positives (TP) which is currently at 98%.

Interactive Knowledge Discovery for Baseline Estimation and Word Segmentation in Handwritten Arabic Text

Electronic document management systems provide great benefits to society. Software tools such as word processors are used in generation, storage, and retrieval of documents in specific formats. Using such tools, documents can be edited, printed, or distributed electronically across networks. However, with paper documents, the previous tasks cannot be accomplished by computers, so there is a need to extract the information in documents to store them in a computerized format. Handwritten text recognition has significant potential.

Real and virtual nature of image edges intersections

Edges intersect in an edge image and this intersection point may correspond to 3D point joining two straight lines in the real world scene and those lines represent a real object plane; in this case it is called a real lines intersection, otherwise it is called a virtual intersection. An automatic system for locating image lines is likely to produce many virtual intersections and so despite many studies in the field of boundary recognition, the question of whether the intersection of two liens in an image of a 3D scene corresponds to a real object point still merits further investigation. This paper presents a computational technique to identify the real or virtual nature of the edge intersections. The discrimination is based on rectified images obtained from a pair of uncalibrated images. The method is tested using different types of images of real scenes. The results obtained showed reliable decisions.