Majlesi Journal of Multimedia Processing
Volume:4 Issue: 1, Mar 2015

Human tracking is an important function for automatic surveillance systems using a vision sensor. Human face is one of the most significant factors to detect person(s) in an image. However, in some cases detection with face alone make false detection especially in single-viewpoints. Therefore, it is difficult to identify a person in an image due to the variety of poses, exactly. This paper describes a method for automatic human tracking based on the nose detection in cropped detected face region using Viola-Jones algorithm and the mean shift tracking method. Additionally, the trackability of this method increases by adjusting the algorithm parameters using different experiments in different condition. The validity of the proposed method is shown through experiment.

An adaptive, scalable, completely blind and robust wavelet-based watermarking approach for color images is proposed. The proposed approach enables scalable watermark detection and provides robustness against progressive wavelet image compression. The binary watermark is divided to three slices and a multiresolution decomposition of each watermark slice is inserted into the selected coefficients of one component image (e.g. red, green and blue component images), adaptively that affect the high activity regions of the image. The watermark insertion is started from the lowest frequency subband of the decomposed component image and each decomposed watermark slice subband is inserted into its counterpart subband of the decomposed component image. In the lowest frequency subband, coefficients with maximum local variance and in the higher frequency subbands, coefficients with maximum magnitude are selected. The watermarked test images are transparent according to the human vision system characteristics and do not show any perceptual degradation. The experimental results very efficiently prove the robustness of the approach against progressive wavelet image coding even at very low bit rates. The watermark extraction process is completely blind and multiple spatial resolutions of the watermark are progressively detectable from the compressed watermarked image. This approach is a suitable candidate for providing efficient authentication for progressive image transmission applications especially over heterogeneous networks, such as the Internet.An adaptive, scalable, completely blind and robust wavelet-based watermarking approach for color images is proposed. The proposed approach enables scalable watermark detection and provides robustness against progressive wavelet image compression. The binary watermark is divided to three slices and a multiresolution decomposition of each watermark slice is inserted into the selected coefficients of one component image (e.g. red, green and blue component images), adaptively that affect the high activity regions of the image. The watermark insertion is started from the lowest frequency subband of the decomposed component image and each decomposed watermark slice subband is inserted into its counterpart subband of the decomposed component image. In the lowest frequency subband, coefficients with maximum local variance and in the higher frequency subbands, coefficients with maximum magnitude are selected. The watermarked test images are transparent according to the human vision system characteristics and do not show any perceptual degradation. The experimental results very efficiently prove the robustness of the approach against progressive wavelet image coding even at very low bit rates. The watermark extraction process is completely blind and multiple spatial resolutions of the watermark are progressively detectable from the compressed watermarked image. This approach is a suitable candidate for providing efficient authentication for progressive image transmission applications especially over heterogeneous networks, such as the Internet.

Funding for training human resources in most countries is very important and costly. Hence in training, prediction of students with expelled risky is one of the today''s key issues and researches. There are imbalances in the training data that causes reduce prediction accuracy in fail students. In this paper, experiments based on data mining techniques have been tried to improve prediction accuracy of fail students. To do this, data from the UCI site are used that contains 5820 records in Turkish students. First, the training data are clustered to select the most appropriate algorithm, Farthest First, and then by doing experiments, best Features including the fitness level of instructor and students'' attendance level and the best Test option, 90% for training data, are selected. Questionnaire is used to Weight features. Finally, the cost-sensitive classification algorithms have been implemented with the proposed cost matrix and the model provided the best results. Results prove that this model can play an important role in promoting science education centers with an accuracy rate of 96.47%, TP rate 99.2% and precision rate of 96%.

In this paper, we studied the segmentation of color images through region merging method. To achieve this goal, images should already be initially segmented and K-means method was used for this task. In region merging method, regions which have the minimum difference in variance merge with each other and this merging continues until the developed regions reach a certain limit. Genetic algorithm method was used to determine that limit. Therefore, merging of the regions repeated several times and each time the number of mutations and crossovers were optimized. The use of this method created very interesting results in terms of identifications. the program segmented crowded images very well and merged regions to a large extent.