Sharif Digital Repository

Image segmentation is middle and an important task in image analysis and machine vision applications. The output images of imaging systems are often fuzzy because of noise, limitation in spatial and temporal resolution, blurring and intensity inhomogeneity in the objects. The goal of this thesis is exploring the fuzzy methods in multispectral image segmentation and proposing a new one to solve some of the recent difficulties and problems. The difficulties and problems such as simultaneous utilization of spatial and spectral information, necessity for dimension reduction, spatial and spectral and intra-cluster image information redundancy, existence of regions with widely varying size,... 

Most broadcast stations insert their logos to claim video content ownership and/or for business purposes.This process reduces picture quality in the logo region and also causes visual discomfort to viewers. Logo detection has a variety of applications; such as automatic TV commercial removal, broadcast surveillance, and video content understanding. Video logos can be considered as a form of visible watermark. Stations may also overlay their logos over the original logo when rebroadcasting a program produced by another station. Also, most comapnies keep a giant archive of their broadcasted programs which are going to be used in their future programs. Here, we have presented an efficient... 

Human face detection is an important research area with several applications such as human computer interface (HCI), face recognition, surveillance systems, security systems, and content-based image retrieval (CBIR). Face detection problem can be stated as “determining whether there are human faces in the image” and if there are “returning the location of each human face in the image” regardless of its position, size, scale, orientation, and lighting condition. In this thesis, we have proposed a new face detection method which combines the AdaBoost algorithm with skin color information and support vector machine (SVM). First, a cascade classifier based on AdaBoost is used to detect faces in... 

In this thesis, change detection in hyperspectral images is investigated. In this process, two hyperspectral images captured from the same scene but in different time instances are given and we intend to detect the occurred changes in the scene. A specific change detection algorithm contains four different steps; namely, preprocessing, selection of the criterion, postprocessing, and decision making. Dimension reduction as a critical process is also deeply investigated to be performed before classification (for decision making purposes.) Two new methods are proposed in the thesis. The proposed MCRD method is designed for dimension reduction and the MPR method is related to the change... 

In this thesis we will first review the problem of deformable modeling. Due to their corss-disciplinary nature, deformable modeling techniques have been the subject of the vigorous research over the past three decades and have found numerous applications in the field of machine vision. Thus the focus will be on general deformable models for computer-based modeling which can be used for computer graphics, visualization, and various image processing applications. So the state of the art of deformable modeling is discussed. Then the focus of the thesis will be on different approaches for the problem of deformable surface reconstruction in 3D space. Amongs these methods there are two main... 

Nowadays, with high performance computers and progress in video recorder devices and their related technologies, price of these devices is being balanced and recording the videos is conventional now. Processing of these videos has been a challenge in industry and science. Control and surveillance of places such as airports and roads and reducing their sizes in memory are some of their research area. Usually one of the initial preprocesses of such applications specially the object tracking is background modeling and background subtraction. This preprocess has an essential effect on the other algorithms that process based on these processes’ results. If this section conducted with fewer... 

Nowadays, wireless communications for multimedia purposes is widely spreaded especially due to video communication demands. One of important considerations in this area is to receive video stream with less effect of errors related to transition channel as well as optimal utilization of bandwidth. The H.264 is exploited as the most efficient video compression standard and because of its predictive nature of encoder and decoder, it is very sensitive to error propagation, so error resilient technics should be applied along with the video compression to reduce the visual impacts of error occurrence and propagation. In this thesis existing methods of video error resilience, video quality measures... 

In many applications of multi view 3D reconstruction, image mosaicing is one of the most important processes. Image mosaicing is the action of stitching some pictures of a scene so that a complete picture of the scene is constructed. The proposed method is such that for PTZ cameras by extracting features from frame, the homography matrix between frames will be computed. Then by converting plane coordinate into spherical coordinate, a complete picture of the scene will be constructed and forground will be removed. For fixed cameras sufficient corresponding points will be selected for computing the homography matrix between cameras. Using the homography, the frames without forground will be... 

In thi¬s thesis we studied 3d reconstruction of synchronized multi-view videos containing dynamic scenes of complicated deformable surfaces and fast motions, such as soccer scenes. Our pupose is to extract shape and motion of scene simultaneousely via a spacio-temporal framework. To this end, we translate our problem to a registration problem both in stereo and temporal pairs. We review 3D reconstruction methods and their related concepts like stereo correspondence and scene flow. This review is necessary to choose the best solution.In proposed solution we formulate 3d recnstruction problem as an optimization problem.Using variational framework we integrate error functional over image planes... 

Estimating 3D pose of the human body in videos has many potential applications in motion analysis, surveillance and human computer interfaces. In the last decade using more than one camera causes more accurate pose estimation. Most of the previous works use calibrated and fixed cameras but in action this is not a realistic assumption and using some arbitrary moving cameras is more likely. We introduce a novel method for this purpose. Our approach does not require specifying locations and directions of the cameras. The method uses a database of 3D poses for different human activities. This database includes body silhouettes from multiple viewpoints. To estimate the 3D body pose, silhouettes... 

Vision-based localization and 3D orientation estimation of a moving camera, has been for long a vast research area including robot localization and mapping, virtual reality and structure from motion. By introduction of RGB-D cameras in 2010, many sparse methods which are based on key-point extraction and tracking, moved towards dense methods. Dense methods utilize the RGB-D depth and gray-scale values in the images and define the odometry estimation problem as an image registration optimization, without the need to make key-point correspondance in images. Although RGB-D cameras impose specific constraints such as limited depth, depth errors and medium resolution, dense methods have shown... 

Today, advances in technology caused the emergence of a need for intelligent systems which can operate without human supervision, and have variety of applications in security, industry, transportation and sport events. Most of these systems require faster and more accurate object tracking. Lots of surveillance and computer vision applications need a process to detect and track objects. For example, the exact location of objects or people is part of the input for applications like movement analysis, counting the number of items, and recognition of people’s behavior. Occlusion is one of the major obstacles in performance of auto-tracking systems. There are several methods to address this... 

The purpose of multi target tracking, is tracking multiple target in each frame of video sequences simultaneously. Object tracking encounters many challenges. Two major challenges of tracking problem that we focus on in them in this thesis are detecting object in each frame of video sequence and handling occlusion problem. Our proposed method for multi target tracking considers one target as a main target and other targets as subsidiary targets. Proposed method track objects in initial view by Kalman filter that is adapted and tuned for multi target tracking. In case of occlusion our method uses data from auxiliary view for handling occlusion. In auxiliary view objects that are occluded in... 

During live broadcast of sports events (such as soccer matches), it is usually useful to provide more information to the audience by adding graphical elements to the video feed; elements such as offside line, distance to goal, and virtual advertisement. This requires to extract information about the camera; for instance, camera position, rotation, and focal length. Hence, we need a method to find the camera calibration data from the image. Previous works have focused on extracting this information based on the points on the pitch. But, that led to insufficient data for projecting 3D points which can be located outside of the pitch. In this thesis, in the preprocessing stage of the first... 

Nowadays TV broadcasting of soccer matches has attracted a lot of viewers. With advancement of technology, the broadcasting is also going under changes. One of these changes is using free-viewpoint video. In free-viewpoint video users can interactively choose any view to watch the match from. Generally free-viewpoint problem is defined as follows. We have sequential frames of a soccer match video from different views as input. We want to create sequential frames of the same video from a different view as output. There are several methods for solving this problem. We can classify these methods into to general approaches. Image-based and geometry-based. We will examine these approaches in... 

During past decade, human pose estimation has become a very important topic in computer vision with many applications; such as robotic, human computer interaction, and analyzing sport videos. Pose estimation is the process of estimating the configuration of the body joints from one or more images. Pose estimation in 2D space has been performed by processing of the images of a single camera. Recently, the multi-view methods have been emerged to estimate 3D human pose.The purpose of this thesis is 3D human pose estimation from multi-view images, captured in football matches. Pictorial structure is an efficient method in 2D human pose estimation. The 3D extension of this method is studied in... 

In the past few years, virtual advertisement replacement has become one of the most popular methods of advertising. Currently a few companies are applying this novel technique on highly-viewd soccer videos. In the atmosphere of rivalry for achieving a method of cheapest and least needed-equpiment between companies, this thesis is aimed at introducing an all software system based on multi-view soccer videos. The goal is to bring the best possible qualities given the challenges of this particular definition. This thesis provides a system to replace advertisement boards around soccer field which consists of detection, tracking and adjusting the advertisement. In detection the systems aligns the... 

Today many videos in an interval of a particular scene in a different view angle to each other are filmed. Use these videos are used in numerous fields. Time synchronization between the video, is the basis for their use. In order to provide an automated process for creating time synchronization between the video, numerous studies have been conducted that each requires specific conditions are to achieve the desired result. The biggest challenge of this problem, which appears when the camera recorded video viewing angles, high and low, and when filming the shared vision between them, jump frames is taken. Players are coordinated. In experiments conducted on different data sets, the average... 

3D reconstruction and tracking of deformable surfaces has received lots of attention in the past decade. This field has been studied from two major aspects: 1) machine vision and 2) computer graphic, and has been used in many applications ranging from industrial design, animation and medical surgeries. One of the most challenging problems in this area is monocular 3D reconstruction, which is an ill-posed problem and involves many ambiguities. Generally, methods which are proposed for monocular 3D reconstruction can be divided into two categories: 1) Template-based methods and 2) Structure from motion methods. Methods of the first category utilize a template model which consists of an image... 

For penalty detection and interpretation of sport match scenes (such as soccer), it is sometimes necessary to watch the scene from an arbitrary viewpoint captured by a virtual camera. In this research, we focus on 3D reconstruction of soccer players using videos that are captured from relatively far distance to the soccer field. In capturing videos from sport scenes, such as soccer scenes, an important issue is the problem of capturing videos from a wide environment to cover almost all parts of the field. Using many cameras with overlapped views might be an answer to this problem but it is impractical due to its cost for capturing and transmitting the data. Furthrmore, because of capturing...