Majlesi Journal of Multimedia Processing
Volume:1 Issue: 4, Dec 2012

Grass field Extraction is most important step to scene analysis, because all important objects in the scene such as players, referee, ball, gate, field line marks, and motion vectors are located on the field. Also the methods are used high-level features such as cinematic features to scene analysis [5,8,9] are based on the amount of grass in the field. This paper presents a new method for grass field Extraction. In the proposed method for extracting the grass, as most existing methods, it is assumed that the soccer field has one distinct dominant colour of green.

This paper proposed a Fuzzy monotonic inclusion (FMI) approach in order to measure similarity between images. Firstly, an image is segmented to several regions, then each region is described by a fuzzy set. Finally, extracted features from each region are mapped into a fuzzy similarity model. FMI scheme makes relation between regions and based on the relations, the regions are selected for the comparison process. Thus, for every image region, both parameters of fuzzy location and area is extracted. We investigated FMI From the conceptual point of view and Semantic relation among Objects. The experimental results on Label Me database, a real world image dataset including 163,000 images, show superiority of the FMI in compared with UFM and Fuzzy Histogram.

As the popularity of digital media is growing, and world is becoming smaller, all due to the internet connectivity and WWW phenomena, the copyright protection of intellectual properties have become a necessity for prevention of illegal copying and content integrity verification. Thus newer data hiding techniques that satisfy the requirements of imperceptibility, robustness, capacity, or data hiding rate and security of the hidden data etc. Are being developed. So we go for digital image watermarking which is a method of authentication data (which is the presence of logo here) embedding in image characteristics with expectation to show resiliency against different type of unintentional or deliberate attacks. Here wavelet transform plays the role of an efficient tool due to its multi-resolution capability. Along with this wavelet transform we mix up another very strong mathematical tool called the singular value decomposition (SVD). Though till date both of them have individually been used as a tool for watermarking of digital media, very few works have utilized their skills in tandem, especially in this area. Our work here by focuses on using both of them together to provide a hybrid technique developed for protection of the intellectual property with better robustness against the popular malicious attacks. This we have seen practically by attacking the watermarked image against simulated attacks and recovering the logo from it.

Co-occurrence matrix is one of the textural features extraction methods which can be used for image morphological classification with acceptable accuracy. This paper deals with textural defects detection and classification algorithm for high-speed steel bar in coil. In the first step, we find the position of defects using co-occurrence matrix and morphology. The extracted suitable features are then presented to a classifier. We use a feedforward backpropagation neural network as a classifier.

Nowadays, automatic suspicious tissue detection in MR images is very important in many diagnostic applications. The diagnosis and separation of cancerous tumors in MR images require accuracy, experience and time, and it has always posed itself as a major challenge to the physicians because they rely on visual detection more than anything in identifying the tumor and determining its location in MR images. In this paper, we have proposed a new method based on image processing techniques that enhance the accuracy, sensitivity and specificity of detection in the diagnosis of cancerous tumors. Initially by applying the DWT on the input images and constructing the approximate coefficients of scaling components, the different parts of image are classified, and with the selection of the appropriate threshold, the suspicious cancerous mass will be separated. The reception of 170 images from the MedPixTm and Harvard Medical School databases including of MRI T1-Wighted, T2-wighted and PD (Proton Density) images which 100 images contain tumor or Edema and other images only represents normal healthy tissue (white and gray tissue) and 97.66% sensitivity and 97.45% accuracy indicate the optimum performance of the system and the correlation coefficient of physician’s early detection with our system were highly significant (p<0.05). The precise positioning of the cancerous tumor and other suspicious tissue enables the physicians to determine the progress level of the disease and recommend the proper treatment proportional to the growth of the cancerous tumor

Least Significant Bit (LSB) is one of the embedding techniques used in Steganography for data hiding purpose. Its simplicity and low computational cost are advantages of this method in comparison with other methods. In this paper, least significant bit method is introduced, reviewed, simulated and realized in hardware. The hardware platform used to implement this method is field programmable gate array (FPGA). Optimized software cores, embedded processor (Micro Blaze) with real-time operating system kernel (XilKernel) and VHDL93 language (for RTL programming style) are used to perform the task as hardware-software co-design. The effects of choosing different embedding layers and presence of different noise sources (pepper and salt, and Gaussian noise) are simulated and discussed. The simulation results show good performance and minimal change in image properties, and hardware realization results show good performance and low resource utilization of this method. The softwares used for simulation and realization are MATLAB, Modelsim, ISE and Leonardo Spectrum.