Sharif Digital Repository

Sparse signal processing (SSP), as a powerful tool and an efficient alternative to traditional complete transforms, has become a focus of attention during the last decade. In this ap-proach, we want to approximate a given signal as a linear combination of as few as possible basis signals. Each basis signal is called an atom and their collection is called a dictionary. This problem is generally difficult and belongs to the NP-hard problems; since it requires a combinatorial search. In recent years however, it has been shown both theoretically and experimentally that the sparset possible representation of a signal in an overcomplete dictio-nary is unique under some conditions and can be found in... 

The sparse decomposition problem or nding sparse solutions of underdetermined linear systems of equations is one of the fundamental issues in signal processing and statistics. In recent years, this issue has been of great interest to researches in various elds of signal processing and accordingly found to be greatly benecial in those elds. This thesis aims at the investigation of the applications of the sparse decomposition problem in image processing. Among dierent applications such as compression, reconstruction, separation and image denoising, this thesis mainly focuses on the last one. One of the methods of image denoising which is closely tied to the sparse decomposition, is the method... 

Recently, sparse channel estimation, i.e. recovering a channel which has much less non zerotaps than its length using a known training sequence, has been a major area of research in the field of sparse signal processing. It can be shown that on the one hand, the underlying unique structure of such channels will make the possibility of estimating the channel taps with the extreme performance, i.e. achieving the Cram´er-Rao bound of the estimation. On the other hand, with an appropriate use of this structure, computational complexity of the receiver (both channel estimator and equalizer) can be reduced by an order. For achieving these goals in this thesis, firstly we have proposed an... 

About a century, “Eye Tracking” has been studied, and it has two definitions: • The process of measuring the point of gaze (where one is looking). • The process of measuring the motion of an eye relative to the head. Eye tracking technology has been used in many fields such as psychology. However, applications of this technology has been recently considered in marketing, computer interfacing, entertainment, training and so forth. Since pupil is a distinc area in eye images, pupil detection is one of the effective solutions of eye tracking. In most of the pupil detection approaches, the edge points of the pupil contour are detected firstly, and then the optimal ellipse is fitted to them.... 

Sparse signal processing (SSP), as a powerful tool and an efficient alternative to traditional complete transforms, has become a focus of attention during the last decade. In this approach, we want to approximate a given signal as a linear combination of as few as possible basis signals. Each basis signal is called an atom and their collection is called a dictionary. This problem is in general difficult and belongs to the Np-hard problems; since it requires a combinatorial search. In recent years however, it has been shown both theoretically and experimentally that the sparset possible representation of a signal in an overcomplete dictionary is unique under some conditions and can be found... 

Sparse representation has attracted much attention over the past decade. The main idea is that natural signals have information contents much lower than their ambient dimensions,and as such, they can be represented by using only a few basis signals (also called atoms). In other words, a natural signal of length n, which in general needs n atoms to be represented, can be written as a linear combination of s atoms, where s ≪ n. To achieve a sparser representation, i.e., a smaller s, the number of atoms is chosen much larger than n. In this way, there are more choices to represent a signal and we can choose the sparsest possible combination. The set of atoms is called a dictionary. Here, two... 

Blind Source Separation (BSS) is a challenging task in signal processing which aims to separate sources from their mixtures when no information is available about the sources or the mixing system. Different approaches have already been proposed for source separation.However, during the last decade, new approaches based on multimodal nature of phenomena have been proposed for source separation. Different aspects of a multimodal phenomenon can be measured by means of different instruments where each of the measured signals is called a modality of that phenomenon. Although the modalities are different signals with different features, due to the same physical origin, they usually have some... 

Sparse representation has been an important problem in recent decade. The main idea in this problem is that natural signals have information contents much lower than their ambient dimensions and as such, they can be represented by using only a few atoms. For example, if the dimension of signal is n, the purpose in sparse representation is to achieve the representation of signal in terms of s atom (s ≪ n). In sparse coding, the dictionary depends on the used signal. In some of the problem, dictionary is specified and sparse representation is obtained by this dictionary. In this case, because the dictionary is known, maybe sparse representation is not suitable for this signal. For this reason,... 

This thesis proposes new algorithms for digital watermarking that are based on Independent Component Analysis (ICA) technique. First, we will show that ICA allows the maximization of the information content and minimization of the induced distortion by decomposing the covertext (in this thesis the image) into statistically independent components. In fact, for a broad class of attacks and fixed capacity values, one can show that distortion is minimized when the message is embedded in statistically independent components. Information theoretical analysis also shows that the information hiding capacity of statistically independent components is maximal. Then we will propose a new wavelet... 

Sparse signal representations and its applications has been a hot topic of research in recent years. It has been demonstrated that sparsity prior can be effectively used as a regularization term to solve many of the inverse problems. One of these problems in which sparse representations have been used is image super-resolution (SR). SR is the problem of finding a high resolution (HR) image from one or several low resolution (LR) images. In this dissertation, we have focused on the problem of finding a HR image from only one LR image which is known as example-based SR. There are two kinds of methods for example-based SR: the methods which use neighborhood embedding and the methods which use... 

Optical Character Recognition is a branch of Image Processing and deals with transforming a scanned document into a text file. The output of a scanning system is in image format and a digital system has no sense about its content. Therefore, this system cannot do any process on its text. For example, words searching, sentences edition and another tasks that are easily done in a word processing software, cannot be directly applied on a scanned document. So, we need a tool to extract the text from a scanned document. A character recognition system consists of different stages: scanning, preprocessing, segmentation, feature extraction, character recognition and post-processing. The purpose of... 

Statistical learning plays a key role in many areas of science [38]. An example of learning problems is image matching, image matching plays an important role in many aspects of computer vision.Computers can be used in intelligent tasks, which are followed by logical inference, for example, visual scenes (images or videos) or speech (audios). For humans visual system of such task are performed hundreds of times every day so easily sometimes without any awareness. In this thesis we focus on the image matching phase which is the first phase of the classification process. One of the popular image matching methods is Scale Invariant Feature Transform (SIFT) which our proposed method is based on... 

Data representation plays an important role in machine learning and the performance of machine learning algorithms for instance, in supervised learnings (e.g. classifcation), and unsupervised ones (e.g. image denoising), are heavily influenced by the input applied to them. Regarding the fact that data usually lacks the desirable quality, efforts are always made to make a more desirable representation of data to be used as input to machine learning algorithms. Among many different representation of data, sparse data representation preserves much more information about data while it is simpler than data. We proposed a new stacked sparse autoencoder by imposing power two of smooth L0 norm of... 

In the past few decades there has been rapid advance in the use of digital cameras in different fields of art and science. Photo editing softwares have provided extensive facilities for their users and graphic softwares have astonished people with artificial yet fabulous images. Under these circumstances, recognition and distinction of authentic images from digitally-manipulated ones have become a critically important but notoriously daunting task. Users of the internet and computers need to recognize authentic images from the manipulated ones, or distinguish a composite photo from an original one. Digital image forensicsa was born as a response to these demands and has so far provided...