This PDF file contains the front matter associated with SPIE Proceedings Volume 9120 including the Title Page, Copyright information, Table of Contents, Introduction, and Conference Committee listing.

In this paper, we study image enhancement technologies for stereoscopic images and an image enhancement algorithm using salient features and wavelet transform is proposed. In the proposed algorithm, the stereoscopic images are decomposed into different subbands and wavelet coefficients are modified based on salient features. Objective and subjective tests were performed to verify the effectiveness of the proposed algorithm. The experimental results show that the proposed algorithm outperforms some conventional algorithms and has great potential in the enhancement of stereoscopic images.

In this paper, we propose an extended pairs of values analysis to detect and estimate the amount of secret messages embedded with 2LSB replacement in digital images based on chi-square attack and regularity rate in pixel values. The detection process is separated from the estimation of the hidden message length, as it is the main requirement of any steganalysis method. Hence, the detection process acts as a discrete classifier, which classifies a given set of images into stego and clean classes. The method can accurately detect 2LSB replacement even when the message length is about 10% of the total capacity, it also reaches its best performance with an accuracy of higher than 0.96 and a true positive rate of more than 0.997 when the amount of data are 20% to 100% of the total capacity. However, the method puts no assumptions neither on the image nor the secret message, as it tested with two sets of 3000 images, compressed and uncompressed, embedded with a random message for each case. This method of detection could also be used as an automated tool to analyse a bulk of images for hidden contents, which could be used by digital forensics analysts in their investigation process.

This paper focuses on steganography based on pixel intensity value decomposition. A number of existing schemes such as binary, Fibonacci, Prime, Natural, Lucas, and Catalan-Fibonacci (CF) are evaluated in terms of payload capacity and stego quality. A new technique based on a specific representation is proposed to decompose pixel intensity values into 16 (virtual) bit-planes suitable for embedding purposes. The proposed decomposition has a desirable property whereby the sum of all bit-planes does not exceed the maximum pixel intensity value, i.e. 255. Experimental results demonstrate that the proposed technique offers an effective compromise between payload capacity and stego quality of existing embedding techniques based on pixel intensity value decomposition. Its capacity is equal to that of binary and Lucas, while it offers a higher capacity than Fibonacci, Prime, Natural, and CF when the secret bits are embedded in 1^st Least Significant Bit (LSB). When the secret bits are embedded in higher bit-planes, i.e., 2^nd LSB to 8^th Most Significant Bit (MSB), the proposed scheme has more capacity than Natural numbers based embedding. However, from the 6^th bit-plane onwards, the proposed scheme offers better stego quality. In general, the proposed decomposition scheme has less effect in terms of quality on pixel value when compared to most existing pixel intensity value decomposition techniques when embedding messages in higher bit-planes.

The Least Significant Bit (LSB) embedding technique is a well-known and broadly employed method in multimedia steganography, used mainly in applications involving single bit-plane manipulations in the spatial domain [1]. The key advantages of LSB procedures are they are simple to understand, easy to implement, have high embedding capacity, and can be resistant to steganalysis attacks. Additionally, the LSB approach has spawned numerous applications and can be used as the basis of more complex techniques for multimedia data embedding. In the last several decades, hundreds of new LSB or LSB variant methods have been developed in an effort to optimize capacity while minimizing detectability, taking advantage of the overall simplicity of this method. LSB-steganalysis research has also intensified in an effort to find new or improved ways to evaluate the performance of this widely used steganographic system. This paper reviews and categorizes some of these major techniques of LSB embedding, focusing specifically on the spatial domain. Some justification for establishing and identifying promising uses of a proposed SD-LSB centric taxonomy are discussed. Specifically, we define a new taxonomy for SD-LSB embedding techniques with the goal of aiding researchers in tool classification methodologies that can lead to advances in the state-of-the-art in steganography. With a common framework to work with, researchers can begin to more concretely identify core tools and common techniques to establish common standards of practice for steganography in general. Finally, we provide a summary on some of the most common LSB embedding techniques followed by a proposed taxonomy standard for steganalysis.

Power centroid radar (PC-Radar) is a fast and powerful adaptive radar scheme that naturally surfaced from the recent discovery of the time-dual for information theory which has been named “latency theory.” Latency theory itself was born from the universal cybernetics duality (UC-Duality), first identified in the late 1970s, that has also delivered a time dual for thermodynamics that has been named “lingerdynamics” and anchors an emerging lifespan theory for biological systems. In this paper the rise of PC-Radar from the UC-Duality is described. The development of PC-Radar, US patented, started with Defense Advanced Research Projects Agency (DARPA) funded research on knowledge-aided (KA) adaptive radar of the last decade. The outstanding signal to interference plus noise ratio (SINR) performance of PC-Radar under severely taxing environmental disturbances will be established. More specifically, it will be seen that the SINR performance of PC-Radar, either KA or knowledgeunaided (KU), approximates that of an optimum KA radar scheme. The explanation for this remarkable result is that PC-Radar inherently arises from the UC-Duality, which advances a “first principles” duality guidance theory for the derivation of synergistic storage-space/computational-time compression solutions. Real-world synthetic aperture radar (SAR) images will be used as prior-knowledge to illustrate these results.

Content-based image retrieval is an automatic process of retrieving images according to image visual contents instead of textual annotations. It has many areas of application from automatic image annotation and archive, image classification and categorization to homeland security and law enforcement. The key issues affecting the performance of such retrieval systems include sensible image features that can effectively capture the right amount of visual contents and suitable similarity measures to find similar and relevant images ranked in a meaningful order. Many different approaches, methods and techniques have been developed as a result of very intensive research in the past two decades. Among many existing approaches, is a cluster-based approach where clustering methods are used to group local feature descriptors into homogeneous regions, and search is conducted by comparing the regions of the query image against those of the stored images. This paper serves as a review of works in this area. The paper will first summarize the existing work reported in the literature and then present the authors’ own investigations in this field. The paper intends to highlight not only achievements made by recent research but also challenges and difficulties still remaining in this area.

In this paper, we propose a mobile system for aiding doctors in skin cancer diagnosis and other persons in skin cancer monitoring. The basic idea is to use image retrieval techniques to help the users to find the similar skin cancer cases stored in a database by using smart phones. The query image can be taken by a smart phone from a patient or can be uploaded from other resources. The shapes of the skin lesions are used for matching two skin lesions, which are segmented from skin images using the skin lesion extraction method developed in 1. The features used in the proposed system are obtained by Fourier descriptor. A prototype application has been developed and can be installed in an iPhone. In this application, the iPhone users can use the iPhone as a diagnosis tool to find the potential skin lesions in a persons’ skin and compare the skin lesions detected by the iPhone with the skin lesions stored in a database in a remote server.

Ultrasound is an effective multipurpose imaging modality that has been widely used for monitoring and diagnosing early pregnancy events. Technology developments coupled with wide public acceptance has made ultrasound an ideal tool for better understanding and diagnosing of early pregnancy. The first measurable signs of an early pregnancy are the geometric characteristics of the Gestational Sac (GS). Currently, the size of the GS is manually estimated from ultrasound images. The manual measurement involves multiple subjective decisions, in which dimensions are taken in three planes to establish what is known as Mean Sac Diameter (MSD). The manual measurement results in inter- and intra-observer variations, which may lead to difficulties in diagnosis. This paper proposes a fully automated diagnosis solution to accurately identify miscarriage cases in the first trimester of pregnancy based on automatic quantification of the MSD. Our study shows a strong positive correlation between the manual and the automatic MSD estimations. Our experimental results based on a dataset of 68 ultrasound images illustrate the effectiveness of the proposed scheme in identifying early miscarriage cases with classification accuracies comparable with those of domain experts using K nearest neighbor classifier on automatically estimated MSDs.

Ethnicity identification of face images is of interest in many areas of application. Different from face recognition of individuals, ethnicity identification classifies faces according to the common features of a specific ethnic group. This paper presents a multi-level fusion scheme for ethnicity identification that combines texture features of local areas of a face using local binary patterns with color features using HSV binning. The scheme fuses the decisions from a k-nearest neighbor classifier and a support vector machine classifier into a final identification decision. We have tested the scheme on a collection of face images from a number of publicly available databases. The results demonstrate the effectiveness of the combined features and improvements on accuracy of identification by the fusion scheme over the identification using individual features and other state-of-art techniques.

Search algorithm complexity modeling, in the form of penetration rate estimation, provides a useful way to estimate search efficiency in application domains which involve searching over a hypothesis space of reference templates or models, as in model-based object recognition, automatic target recognition, and biometric recognition. The penetration rate quantifies the expected portion of the database that must be searched, and is useful for estimating search algorithm computational requirements. In this paper we perform mathematical modeling to derive general equations for penetration rate estimates that are applicable to a wide range of recognition problems. We extend previous penetration rate analyses to use more general probabilistic modeling assumptions. In particular we provide penetration rate equations within the framework of a model-based image alignment application domain in which a prioritized hierarchical grid search is used to rank subspace bins based on matching probability. We derive general equations, and provide special cases based on simplifying assumptions. We show how previously-derived penetration rate equations are special cases of the general formulation. We apply the analysis to model-based logo image alignment in which a hierarchical grid search is used over a geometric misalignment transform hypothesis space. We present numerical results validating the modeling assumptions and derived formulation.

Advanced Kalman Filters has been used extensively in the domain of video based tracking of target objects. They can be viewed as an extension of Kalman Filtering principle. Instead of using object point mass as a tracker as used in the Kalman filter, alterations are made to incorporate advanced strategies. This is the typical formulation of the Kalman Enhanced Filter (KEF). Even though this allows the use of non-linearity for state prediction, it is constrained by its choice of the Kalman state transition function. Furthermore the KEF does not provide a methodology of selection of the distribution of the prior. The proper tuning of the above choices is critical for performance of the KEF. This work addresses these constraints of the KEF. It particularly targets two significant areas. Firstly it automates the state matrix generation process by fusing alternate tracking mechanism to the KEF. This novel technique is tested for tracking of real video sequence and its efficacy is quantified.

Face classification of multiple cameras has wide applications in surveillance. In this paper, the efficacy of a multi-frame decision-level fusion scheme for face classification based on the photon-counting linear discriminant analysis is investigated. The photon-counting linear discriminant analysis method is able to realize Fisher’s criterion without preprocessing for dimensionality reduction. The decision-level fusion scheme is comprised of three stages: score normalization, score validation, and score combination. After normalization, the candidate scores are selected and combined by means of a score validation process and a fusion rule, respectively, in order to generate a final score. In the experiments, out-of-focus and motion blurs are rendered on test images simulating harsh conditions.

In this paper, we propose an analytic sequential methods for detecting port-scan attackers which routinely perform random “portscans” of IP addresses to find vulnerable servers to compromise. In addition to rigorously control the probability of falsely implicating benign remote hosts as malicious, our method performs significantly faster than other current solutions. We have developed explicit formulae for quick determination of the parameters of the new detection algorithm.

Video compression and encryption became very essential in a secured real time video transmission. Applying both techniques simultaneously is one of the challenges where the size and the quality are important in multimedia transmission. In this paper we proposed a new technique for video compression and encryption. Both encryption and compression are based on edges extracted from the high frequency sub-bands of wavelet decomposition. The compression algorithm based on hybrid of: discrete wavelet transforms, discrete cosine transform, vector quantization, wavelet based edge detection, and phase sensing. The compression encoding algorithm treats the video reference and non-reference frames in two different ways. The encryption algorithm utilized A5 cipher combined with chaotic logistic map to encrypt the significant parameters and wavelet coefficients. Both algorithms can be applied simultaneously after applying the discrete wavelet transform on each individual frame. Experimental results show that the proposed algorithms have the following features: high compression, acceptable quality, and resistance to the statistical and bruteforce attack with low computational processing.

In the world of computer and network security, there are myriad ways to launch an attack, which, from the perspective of a network, can usually be defined as "traffic that has huge malicious intent." Firewall acts as one of the measure in order to secure the device from incoming unauthorized data. There are infinite number of computer attacks that no firewall can prevent, such as those executed locally on the machine by a malicious user. From the network's perspective, there are numerous types of attack. All the attacks that degrade the effectiveness of data can be grouped into two types: brute force and precision. The Firewall that belongs to Juniper has the capability to protect against both types of attack. Denial of Service (DoS) attacks are one of the most well-known network security threats under brute force attacks, which is largely due to the high-profile way in which they can affect networks. Over the years, some of the largest, most respected Internet sites have been effectively taken offline by Denial of Service (DOS) attacks. A DoS attack typically has a singular focus, namely, to cause the services running on a particular host or network to become unavailable. Some DoS attacks exploit vulnerabilities in an operating system and cause it to crash, such as the infamous Win nuke attack. Others submerge a network or device with traffic so that there are no more resources to handle legitimate traffic. Precision attacks typically involve multiple phases and often involves a bit more thought than brute force attacks, all the way from reconnaissance to machine ownership. Before a precision attack is launched, information about the victim needs to be gathered. This information gathering typically takes the form of various types of scans to determine available hosts, networks, and ports. The hosts available on a network can be determined by ping sweeps. The available ports on a machine can be located by port scans. Screens cover a wide variety of attack traffic as they are configured on a per-zone basis. Depending on the type of screen being configured, there may be additional settings beyond simply blocking the traffic. Attack prevention is also a native function of any firewall. Juniper Firewall handles traffic on a per-flow basis. We can use flows or sessions as a way to determine whether traffic attempting to traverse the firewall is legitimate. We control the state-checking components resident in Juniper Firewall by configuring "flow" settings. These settings allow you to configure state checking for various conditions on the device. You can use flow settings to protect against TCP hijacking, and to generally ensure that the fire-wall is performing full state processing when desired. We take a case study of attack on a network and perform study of the detection of the malicious packets on a Net screen Firewall. A new solution for securing enterprise networks will be developed here.

As speech based operation becomes a main hand-free interaction solution between human and mobile devices (i.e., smartphones, Google Glass), privacy preserving speaker verification receives much attention nowadays. Privacy preserving speaker verification can be achieved through many different ways, such as fuzzy vault and encryption. Encryption based solutions are promising as cryptography is based on solid mathematic foundations and the security properties can be easily analyzed in a well established framework. Most current asymmetric encryption schemes work on finite algebraic structures, such as finite group and finite fields. However, the encryption scheme for privacy preserving speaker verification must handle floating point numbers. This gap must be filled to make the overall scheme practical. In this paper, we propose a number system that meets the requirements of both speaker verification and the encryption scheme used in the process. It also supports addition homomorphic property of Pailliers encryption, which is crucial for privacy preserving speaker verification. As asymmetric encryption is expensive, we propose a method of packing several numbers into one plain-text and the computation overhead is greatly reduced. To evaluate the performance of this method, we implement Pailliers encryption scheme over proposed number system and the packing technique. Our findings show that the proposed solution can fulfill the gap between speaker verification and encryption scheme very well, and the packing technique improves the overall performance. Furthermore, our solution is a building block of encryption based privacy preserving speaker verification, the privacy protection and accuracy rate are not affected.

Omni-directional high resolution surveillance has a wide application range in defense and security fields. Early systems used for this purpose are based on parabolic mirror or fisheye lens where distortion due to the nature of the optical elements cannot be avoided. Moreover, in such systems, the image resolution is limited to a single image sensor’s image resolution. Recently, the Panoptic camera approach that mimics the eyes of flying insects using multiple imagers has been presented. This approach features a novel solution for constructing a spherically arranged wide FOV plenoptic imaging system where the omni-directional image quality is limited by low-end sensors. In this paper, an overview of current Panoptic camera designs is provided. New results for a very-high resolution visible spectrum imaging and recording system inspired from the Panoptic approach are presented. The GigaEye-1 system, with 44 single cameras and 22 FPGAs, is capable of recording omni-directional video in a 360°×100° FOV at 9.5 fps with a resolution over (17,700×4,650) pixels (82.3MP). Real-time video capturing capability is also verified at 30 fps for a resolution over (9,000×2,400) pixels (21.6MP). The next generation system with significantly higher resolution and real-time processing capacity, called GigaEye-2, is currently under development. The important capacity of GigaEye-1 opens the door to various post-processing techniques in surveillance domain such as large perimeter object tracking, very-high resolution depth map estimation and high dynamicrange imaging which are beyond standard stitching and panorama generation methods.

Image or video inpainting is the process/art of retrieving missing portions of an image without introducing undesirable artifacts that are undetectable by an ordinary observer. An image/video can be damaged due to a variety of factors, such as deterioration due to scratches, laser dazzling effects, wear and tear, dust spots, loss of data when transmitted through a channel, etc. Applications of inpainting include image restoration (removing laser dazzling effects, dust spots, date, text, time, etc.), image synthesis (texture synthesis), completing panoramas, image coding, wireless transmission (recovery of the missing blocks), digital culture protection, image de-noising, fingerprint recognition, and film special effects and production. Most inpainting methods can be classified in two key groups: global and local methods. Global methods are used for generating large image regions from samples while local methods are used for filling in small image gaps. Each method has its own advantages and limitations. For example, the global inpainting methods perform well on textured image retrieval, whereas the classical local methods perform poorly. In addition, some of the techniques are computationally intensive; exceeding the capabilities of most currently used mobile devices. In general, the inpainting algorithms are not suitable for the wireless environment. This paper presents a new and efficient scheme that combines the advantages of both local and global methods into a single algorithm. Particularly, it introduces a blind inpainting model to solve the above problems by adaptively selecting support area for the inpainting scheme. The proposed method is applied to various challenging image restoration tasks, including recovering old photos, recovering missing data on real and synthetic images, and recovering the specular reflections in endoscopic images. A number of computer simulations demonstrate the effectiveness of our scheme and also illustrate the main properties and implementation steps of the presented algorithm. Furthermore, the simulation results show that the presented method is among the state-of-the-art and compares favorably against many available methods in the wireless environment. Robustness in the wireless environment with respect to the shape of the manually selected “marked” region is also illustrated. Currently, we are working on the expansion of this work to video and 3-D data.

In this paper, we introduce the mobile embedded system implemented for capturing stereo image based on two CMOS camera module. We use WinCE as an operating system and capture the stereo image by using device driver for CMOS camera interface and Direct Draw API functions. We send the raw captured image data to the host computer by using WiFi wireless communication and then use GPU hardware and CUDA programming for implementation of real time three-dimensional stereo image by synthesizing the depth of ROI(region of interest). We also try to find and declare the mechanism of deblurring of CMOS camera module based on the Kirchhoff diffraction formula and propose a deblurring model. Synthesized stereo image is real time monitored on the shutter glass type three-dimensional LCD monitor and disparity values of each segment are analyzed to prove the validness of emphasizing effect of ROI.

Security and surveillance videos, due to usage in open environments, are likely subjected to low resolution, underexposed, and overexposed conditions that reduce the amount of useful details available in the collected images. We propose an approach to improve the image quality of low resolution images captured in extreme lighting conditions to obtain useful details for various security applications. This technique is composed of a combination of a nonlinear intensity enhancement process and a single image super resolution process that will provide higher resolution and better visibility. The nonlinear intensity enhancement process consists of dynamic range compression, contrast enhancement, and color restoration processes. The dynamic range compression is performed by a locally tuned inverse sine nonlinear function to provide various nonlinear curves based on neighborhood information. A contrast enhancement technique is used to obtain sufficient contrast and a nonlinear color restoration process is used to restore color from the enhanced intensity image. The single image super resolution process is performed in the phase space, and consists of defining neighborhood characteristics of each pixel to estimate the interpolated pixels in the high resolution image. The combination of these approaches shows promising experimental results that indicate an improvement in visibility and an increase in usable details. In addition, the process is demonstrated to improve tracking applications. A quantitative evaluation is performed to show an increase in image features from Harris corner detection and improved statistics of visual representation. A quantitative evaluation is also performed on Kalman tracking results.

This paper concerns available steganographic techniques that can be used for sending hidden data through public network. Typically, in steganographic communication it is advised to use popular/often used method for sending hidden data and amount of that data need to be high as much as possible. We confirmed this by choosing a Domain Name System (DNS) as a vital protocol of each network and choosing Distributed denial of service (DDoS) attacks that are most popular network attacks currently represented in the world. Apart from characterizing existing steganographic methods we provide new insights by presenting two new techniques. The first one is network steganography solution which exploits free/unused protocols fields and is known for IP, UDP or TCP protocols, but has never been applied to DNS (Domain Name Server) which are the fundamental part of network communications. The second explains the usage of DNS Amplification DDoS Attack to send seamlessly data through public network. The calculation that was performed to estimate the total amount of data that can be covertly transferred by using these technique, regardless of steganalysis, is included in this paper.

This paper focuses on a machine learning approach for objective inpainting quality assessment. Inpainting has received a lot of attention in recent years and quality assessment is an important task to evaluate different image reconstruction approaches. Quantitative metrics for successful image inpainting currently do not exist; researchers instead are relying upon qualitative human comparisons in order to evaluate their methodologies and techniques. We present an approach for objective inpainting quality assessment based on natural image statistics and machine learning techniques. Our method is based on observation that when images are properly normalized or transferred to a transform domain, local descriptors can be modeled by some parametric distributions. The shapes of these distributions are different for noninpainted and inpainted images. Approach permits to obtain a feature vector strongly correlated with a subjective image perception by a human visual system. Next, we use a support vector regression learned on assessed by human images to predict perceived quality of inpainted images. We demonstrate how our predicted quality value repeatably correlates with a qualitative opinion in a human observer study.

Color image quality measures have been used for many computer vision tasks. In practical applications, the no-reference (NR) measures are desirable because reference images are not always accessible. However, only limited success has been achieved. Most existing NR quality assessments require that the types of image distortion is known a-priori. In this paper, three NR color image attributes: colorfulness, sharpness and contrast are quantified by new metrics. Using these metrics, a new Color Quality Measure (CQM), which is based on the linear combination of these three color image attributes, is presented. We evaluated the performance of several state-of-the-art no-reference measures for comparison purposes. Experimental results demonstrate the CQM correlates well with evaluations obtained from human observers and it operates in real time. The results also show that the presented CQM outperforms previous works with respect to ranking image quality among images containing the same or different contents. Finally, the performance of CQM is independent of distortion types, which is demonstrated in the experimental results.

The paper deals with presentation of the IVAS system within the 7FP EU INDECT project. The INDECT project aims at developing the tools for enhancing the security of citizens and protecting the confidentiality of recorded and stored information. It is a part of the Seventh Framework Programme of European Union. We participate in INDECT portal and the Interactive Video Audio System (IVAS). This IVAS system provides a communication gateway between police officers working in dispatching centre and police officers in terrain. The officers in dispatching centre have capabilities to obtain information about all online police officers in terrain, they can command officers in terrain via text messages, voice or video calls and they are able to manage multimedia files from CCTV cameras or other sources, which can be interesting for officers in terrain. The police officers in terrain are equipped by smartphones or tablets. Besides common communication, they can reach pictures or videos sent by commander in office and they can respond to the command via text or multimedia messages taken by their devices. Our IVAS system is unique because we are developing it according to the special requirements from the Police of the Czech Republic. The IVAS communication system is designed to use modern Voice over Internet Protocol (VoIP) services. The whole solution is based on open source software including linux and android operating systems. The technical details of our solution are presented in the paper.

This paper proposes a novel texture descriptor based indices of degrees of local approximating polynomials. An input image is divided into non-overlapping patches which are reshaped into a one-dimensional source vectors. These vectors are approximated using local polynomial functions of various degrees. For each element of the source vector, these approximations are ranked according to the difference between the original and approximated values. A set of indices of polynomial degrees form a local feature. This procedure is repeated for every pixel. Finally, a proposed texture descriptor is obtained from the frequency histogram of all obtained local features. A nearest neighbor classifier utilizing distance metric is used to evaluate a performance of the introduced descriptor on the following datasets: Brodatz, KTH-TIPS, KTH-TIPS2b, UCLA and Columbia-Utrecht (CUReT) with respect to different methods of texture analysis and classification. A proper parameter setup of the proposed texture descriptor is discussed. The results of this comparison demonstrate that the proposed method is competitive with the recent statistical approaches such as local binary patterns (LBP), local ternary patterns, completed LBP, Weber’s local descriptor, and VZ algorithms (VZ-MR8 and VZ-Joint). At the same time, on KTH-TIPS2-b and KTH-TIPS datasets, the proposed method is slightly inferior to some of the state-of-the-art methods.

Image alignment and mosaicing are usually performed on a set of overlapping images, using features in the area of overlap for seamless stitching. In many cases such images have different size and shape. So we need to crop panoramas or to use image extrapolation for them. This paper focuses on novel image inpainting method based on modified exemplar-based technique. The basic idea is to find an example (patch) from an image using local binary patterns, and replacing non-existed (‘lost’) data with it. We propose to use multiple criteria for a patch similarity search since often in practice existed exemplar-based methods produce unsatisfactory results. The criteria for searching the best matching uses several terms, including Euclidean metric for pixel brightness and Chi-squared histogram matching distance for local binary patterns. A combined use of textural geometric characteristics together with color information allows to get more informative description of the patches. In particular, we show how to apply this strategy for image extrapolation for photo stitching. Several examples considered in this paper show the effectiveness of the proposed approach on several test images.