Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

MUCKE aims to mine a large volume of images, to structure them conceptually and to use this conceptual structuring in order to improve large-scale image retrieval. The last decade witnessed important progress concerning low-level image representations. However, there are a number problems which need to be solved in order to unleash the full potential of image mining in applications. The central problem with low-level representations is the mismatch between them and the human interpretation of image content. This problem can be instantiated, for instance, by the incapability of existing descriptors to capture spatial relationships between the concepts represented or by their incapability to convey an explanation of why two images are similar in a content-based image retrieval framework. We start by assessing existing local descriptors for image classification and by proposing to use co-occurrence matrices to better capture spatial relationships in images. The main focus in MUCKE is on cleaning large scale Web image corpora and on proposing image representations which are closer to the human interpretation of images. Consequently, we introduce methods which tackle these two problems and compare results to state of the art methods. Note: some aspects of this deliverable are withheld at this time as they are pending review. Please contact the authors for a preview.

http://ieeexplore.ieee.org/iel2/4524/12820/00592460.pdf

Image Processing

A comprehensive survey on support vector machine classification: Applications, challenges and trends

Texture features for image classification and retrieval.

In this paper, an image accreditation technique by embedding digital watermarks in images is proposed. The proposed method for the digital watermarking is based on the wavelet transform. This is unlike most previous work, which used a random number of a sequence of bits as a watermark and where the watermark can only be detected by comparing an experimental threshold value to determine whether a sequence of random signals is the watermark. The proposed approach embeds a watermark with visual recognizable patterns, such as binary, gray, or color image in images by modifying the frequency part of the images. In the proposed approach, an original image is decomposed into wavelet coefficients. Then, multi-energy watermarking scheme based on the qualified significant wavelet tree (QSWT) is used to achieve the robustness of the watermarking. Unlike other watermarking techniques that use a single casting energy, QSWT adopts adaptive casting energy in different resolutions. The performance of the proposed watermarking is robust to a variety of signal distortions, such as JPEG, image cropping, sharpening, median filtering, and incorporating attacks.

Hiding digital watermarks using multiresolution wavelet transform

A new watermarking scheme which incorporates wavelet and spatial transformation has been developed for digital images. This algorithm utilizes the wavelet multiresolutional structure to construct the image frequency components and the chaotic transformation as two dimensional integer vector generators for spatial transform to select the location during the watermark embedding. To efficiently embed the watermark within the image without the loss of image quality and provide the robustness for the watermark detection under attacks, a modular based spatial threshold and adjustment scheme of the wavelet coefficients has been developed in this research. Unlike other watermarking schemes which usually rely on significantly large amount of side information for watermark detection, our algorithm need only few key parameters to detect the watermark with meaningful content. Compared with some known approaches of watermarking schemes, this algorithm results in superior robustness and information protection for keeping watermark intact under image processing attacks like image compression.

Joint wavelet and spatial transformation for digital watermarking

Digital forensics has lately become one of the very important applications to identify the characteristics and the originality of the digital devices. This study has focused on analyzing the relationship between digital cameras and the photographs by using the support vector machine (SVM). Based on the fact that the internal imaging formation algorithms of the cameras are different from one manufacturer to another, our approach first calculates the characteristic values of the images taken by different cameras in conjunction with image processing techniques and data exploration methods. The training and categorization procedures of the image characteristic values are further conducted through SVM to identify the source camera of the images. Based on SVM's ability to distinguish cameras of different brands, this study also examines whether the method can differentiate cameras of the same brand, or even the popular mobile phones with camera. The experiment results demonstrate that our approach can achieve higher identification rate for camera and mobile phone sources than the results from other literatures.

Camera/Mobile Phone Source Identification for Digital Forensics

A general watermarking scheme which adopts the multiresolution wavelet packet transform and content based watermarking scheme has been developed for a digital image. The systematic approach of the watermarking includes the selection of the image content, the wavelet packet transform and the implementation of total automorphism as the spatial transform for the watermark. To efficiently embed the watermark within the images and provide the robustness for the watermark detection under attacks, a modular based element classification and adjustment of the wavelet transformed coefficients has been utilized in this research. Instead of using the random number generator to create the watermark, the meaningful and recognizable seal image has been used as the watermark which provides immediately strong authentication information. In addition, the parameter settings of the choices among the transforms serve as the key information in deciphering an watermarked image without referring to the original image. Compared with other watermarking methods, our technique provides a generalized approach for watermarking with robustness.

Wavelet packet and adaptive spatial transformation of watermark for digital image authentication

Recently, digital forensics, which involves the collection and analysis of the origin digital device, has become an important issue. Digital content can play a crucial role in identifying the source device, such as serve as evidence in court. To achieve this goal, we use different texture feature extraction methods such as graylevel co-occurrence matrix (GLCM) and discrete wavelet transform (DWT), to analyze the Chinese printed source in order to find the impact of different output devices. Furthermore, we also explore the optimum feature subset by using feature selection techniques and use support vector machine (SVM) to identify the source model of the documents. The average experimental results attain a 98.64 % identification rate which is significantly superior to the existing known method of GLCM by 1.27 %. The superior testing performance demonstrates that the proposed identification method is very useful for source laser printer identification.

Digital forensics of printed source identification for Chinese characters

Digital forensics has lately become one of the very important applications to identify the characteristics and the originality of the digital devices. This study has focused on analyzing the relationship between digital cameras and the associated images. Digital image processing technology has been applied with data mining method to get images' features. Those features were trained and classified to identify the camera sources from the images. In addition, the identification features were further categorized for improvements, based on the analyses, to enhance the precision rate of the identification. This research also compared not only cameras of different brand, but also those of the same manufacturer with similar models. This research has found that the feature based approach has better performance to distinguish the camera sources among brands. Further findings are discussed and suggested for the potential limit of the identification methods in real applications.

Min-Jen Tsai

Papers

Joint wavelet and spatial transformation for digital watermarking

Camera/Mobile Phone Source Identification for Digital Forensics

Wavelet packet and adaptive spatial transformation of watermark for digital image authentication

Digital forensics of printed source identification for Chinese characters

USING Image Features to Identify Camera Sources