Majlesi Journal of Multimedia Processing
Volume:3 Issue: 2, Jun 2014

Edge detection is a crucial preparatory stage in image processing. The current edge detection methods suffer from the problem of high frailty to noise. In the present paper, a new method for image edge detection in images damaged by salt-and-pepper noise introduced. The simple structure of the proposed method is composed of four neural networks, a neuro-fuzzy network and an adaptive median filter. The internal parameters of these networks are adaptively optimized in training through using simple synthetic images that can be generated in a computer. The proposed method is tested on many popular images such as cameraman,boat,lena and etc. Then the results have been compared with those of the previous edge detectors such as Sobel and Canny. Empirical statistics show that the newly-introduced method presents much better performance than the previous ones and can be benefited in any process of edge detection of the Salt-and-pepper noise-damaged images.

In this paper, a new model has been introduced for automatic text classification. This model includes two phases: text pre-processing and text classification. Pre-processing phase includes text preparation, indexing and indices weighting phases. Preparation phase of suggested model includes getting the words root, prefixes and suffixes omission, stop words omission and writing marks. In the proposed model, we used N-gram method for indexing in pre-processing phase and in order to weight the indices, TF.IDF filter was applied. In the second phase, K-Nearest Neighbor (K-NN) was used to train the model for classifying. In order to evaluate and compare the results, precision and recall measures; Micro – F1 & Macro – F1 have been calculated for different methods of indexing. The results of the tests conducted on Reuters 21578 data set showed that our proposed method has the best efficiency to Naïve Bayes & Decision Tree algorithms on this data set.

Many classes of images contain spatial regions which are more important than other regions. Compression methods capable of delivering higher reconstruction quality for important parts are attractive in this situation. For medical images, only a small portion of the image might be diagnostically useful, but the cost of a wrong interpretation is high. Lossless compression schemes with secure transmission play a key role in telemedicine applications that help in accurate diagnosis and research. In this paper, we propose lossless compression method for Digital Imaging and Communications in Medicine images. The method includes the compression of region of interest using lossless image compression technique i.e. Huffman coding while the remaining irrelevant area of image (background) is compressed using the near lossless image compression techniques i.e. SPIHT. The main objective of this work is to reject the noisy background and reconstruct the image portions in a lossless manner. The image later reconstructed by merging the region of interest and background. All the combinations are compared and minor difference in their performances is observed.

Voice recognition is identifying someone’s Speech. It is usually done by signal processing and extracting speech information. In this paper an intelligent method has been implemented to decrease the consuming time for analysis computation. The MLP (Multilayer Perceptron) technique with back propagation learning rule is employed for a couple of speakers whose speech content has not been trained for network previously

One of the most important methods for the measurement of subcutaneous body fat is in the field of clinical MRI, in this way the desired area of fat tissue (ROI) manually isolated and the region of interest identified by the radiologist, and its thickness is measured. But this separation is not necessary is has not the required speed and accuracy. In contrast, the method presented in this paper is an automated method for fast and accurate measurement of subcutaneous fat thickness. The recent study is done on 100 MRI images of (50 female, 50 male), and the thickness of subcutaneous fat in the midline of the abdomen and flank (from the iliac crest to the rib area) based on the proposed method (K-MEANS and FCM clustering methods and using recursive connected components algorithm) were calculated.