Papers by Alexander Sibiryakov
… Vision and Pattern Recognition (CVPR), 2011 …, Jan 1, 2011
This paper proposes a new template matching method that is robust to outliers and fast enough for... more This paper proposes a new template matching method that is robust to outliers and fast enough for real-time operation. The template and image are densely transformed in binary code form by projecting and quantizing histograms of oriented gradients. The binary codes are matched by a generic method of robust similarity applicable to additive match measures, such as L p -and Hamming distances. The robust similarity map is computed efficiently via a proposed Inverted Location Index structure that stores pixel locations indexed by their values. The method is experimentally justified in large image patch datasets. Challenging applications, such as intra-category object detection, object tracking, and multimodal image matching are demonstrated.
… , 2007. ICCE 2007. Digest of Technical …, Jan 1, 2007
… Electronics, 2008. ICCE 2008. Digest of …, Jan 1, 2008
We present a very low-complexity method for estimation of parametric global motion in video. Firs... more We present a very low-complexity method for estimation of parametric global motion in video. First, the motion is approximated by translation, then small peripheral blocks from the rectangular area of overlap are matched, and local motion vectors (MVs) are extracted. Both global translation and local MVs are efficiently computed using 1D phase correlation of integral image projections. The global parametric motion is estimated from inlier MVs found by the RANSAC. Our main contributions, enabling high performance, are: 1) non-informative block rejection based on image projection analysis, 2) informative block search using line scanning from peripheral blocks towards the central block, and 3) iterative replacement of the outlier MVs. The achieved performance is ~1000fps for CIF frame. The method is demonstrated in video sequences with rapid camera motion, zoom and rotation. Geometric and photometric comparisons with two other motion estimation methods are performed.
… (ICCE), 2011 IEEE International Conference on, Jan 1, 2011
The hand jitter usually has a negative effect on recorded video, but combined with intentional ca... more The hand jitter usually has a negative effect on recorded video, but combined with intentional camera motion, it forms a unique global motion pattern intrinsic to the video sequence. Therefore a useful application that involves hand jitter would be to identify videos taken by hand-held devices by extracting and matching a compact descriptor entirely based on global motion. We confirm this idea via a video matching experiment that shows how even a simple and fast global motion extraction method, based on integral projections, provides excellent short clip identification and localization results, despite severe video modifications.
WSEAS Transactions on …, Jan 1, 2005
Proceedings of the 5th WSEAS …, Jan 1, 2005
This paper addresses a problem of robust, accurate and fast object detection in complex environme... more This paper addresses a problem of robust, accurate and fast object detection in complex environments, such as cluttered backgrounds and low-quality images. A new method called Statistical Template Matching is proposed to detect objects, represented by a set of template regions. A similarity measure between the image and a template is derived from the Fisher criterion. We show how to apply our method to face and facial feature detection tasks, and demonstrate its performance in some difficult cases, such as moderate variation of scale factor of the object, local image warping and distortions caused by image compression.
Optical 3D Measurement Techniques, Zurich, …, Jan 1, 1997
Precise points matching on the images of a stereopair is one of central problems in the area of m... more Precise points matching on the images of a stereopair is one of central problems in the area of machine vision and digital photogrammetry. A lot of publications is devoted to investigations of this problem. Among well-known classical approaches the conventional normalized cross-correlation method occupies first place due to its fundamental importance and vast utilizing in practice during several decades. However, revealing drawbacks of the method connected with non-adaptive geometric properties have brought the creating of new more powerful methods, for example, adaptive least squares correlation . The goal of this article is to provide consequential extension of classical normal cross-correlation that it could gain subpixel accuracy and adaptive geometric properties. It is shown that consecutive normalized cross-correlation application results in problem of a finding a vector of the amendments of six-parameter affine transformation as a generalized eigenvector problem. The theoretical decision of this problem in view of specific structure of matrixes obtained by linearization is offered. Effective algorithm of the numerical solution based on a triangular Cholecky decomposition is proposed. The main result obtained in the present paper is the proof of equivalence of least-square correlation and an adaptive extension of cross-correlation.
Discrete Geometry for Computer Imagery, Jan 1, 2008
We present a novel template matching framework for detecting geometrically transformed objects. A... more We present a novel template matching framework for detecting geometrically transformed objects. A template is a simplified representation of the object of interest by a set of pixel groups of any shape, and the similarity between the template and an image region is derived from the F-test statistic. The method selects a geometric transformation from a discrete set of transformations, giving the best statistical independence of such groups Efficient matching is achieved using 1D analogue of integral images -integral lines, and the number of operations required to compute the matching score is linear with template size, comparing to quadratic dependency in conventional template matching. Although the assumption that the geometric deformation can be approximated from discrete set of transforms is restrictive, we introduce an adaptive subpixel refinement stage for accurate matching of object under arbitrary parametric 2D-transformation. The parameters maximizing the matching score are found by solving an equivalent eigenvalue problem. The methods are demonstrated on synthetic and real-world examples and compared to standard template matching methods.
Lecture Notes in Computer Science 2897 Edited by G. Goos, J. Hartmanis, and J. van Leeuwen This O... more Lecture Notes in Computer Science 2897 Edited by G. Goos, J. Hartmanis, and J. van Leeuwen This One TTK3-4JA-75HF ... Springer Berlin Heidelberg New York Hong Kong London Milan Paris Tokyo ... Olivier Balet Gerard Subsol Patrice Torguet (Eds.) Virtual Storytelling ...
… Imaging V. Edited by C. Bouman, …, Jan 1, 2007
This paper presents a multi-image registration method, which aims at recognizing and extracting m... more This paper presents a multi-image registration method, which aims at recognizing and extracting multiple panoramas from an unordered set of images without user input. A method for panorama recognition introduced by Lowe and Brown [1] is based on extraction of a full set of scale invariant image features and fast matching in feature space, followed by post-processing procedures. We propose a different approach, where the full set of descriptors is not required, and a small number of them are used to register a pair of images. We propose feature point indexing based on corner strength value. By matching descriptor pairs with similar corner strengths we update clusters in rotation-scale accumulators, and a probabilistic approach determines when these clusters are further processed with RANSAC to find inliers of image homography. If the number of inliers and global similarity between images are sufficient, a fast geometry-guided point matching is performed to improve the accuracy of registration. A global registration graph, whose node weights are proportional to the image similarity in the area of overlap, is updated with each new registration. This allows the prediction of undiscovered image registrations by finding the shortest paths and corresponding transformation chains. We demonstrate our approach using typical image collections containing multiple panoramic sequences.
Multimedia Signal Processing, 2009. MMSP'09. …, Jan 1, 2009
We present a method of transforming local image descriptors into a compact form of bit-sequences ... more We present a method of transforming local image descriptors into a compact form of bit-sequences whose similarity is determined by Hamming distance. Following the Locality-Sensitive Hashing approach, the descriptors are projected on a set of random directions that are learned from a set of non-matching data. The learned random projections result in high-entropy binary codes (HE 2 ) that outperform codes based on standard random projections in match/non-match classification and nearest neighbor search. Despite of data compression and granularity of Hamming space, HE 2 -descriptor outperforms the original descriptor in the classification task. In nearest neighbor search task, the performance of the HE 2 -descriptor is asymptotic to the performance of the original descriptor. As a supporting result, we obtain another descriptor, HE 2 +1, and demonstrate that the performance of the original descriptor can be improved by adding a few bits derived from the descriptor itself.
… Workshop on Automatic …, Jan 1, 2001
The modern situation in Russia in the field of geo-informational support of decision making is ch... more The modern situation in Russia in the field of geo-informational support of decision making is characterized by lack of regular high fidelity original data on the one hand and the predominance of low cost computer maintenance on the other hand. This paper is devoted to elements of the technology developed in GosNIIAS aimed at using of various data sources on PC-level computers to provide 3D site modeling for a wide variety of applications, from military mission planning, mission rehearsal and reconnaissance data analysis to urban and landscape planning, transport flow planning and the virtual tourism .
Internal report, Institute of Geodesy and …, Jan 1, 1996
INTERNATIONAL ARCHIVES OF …, Jan 1, 2000
The paper describes a number of methods for fast DTM generation, processing and merging implement... more The paper describes a number of methods for fast DTM generation, processing and merging implemented in the new powerful PC-based photogrammetric system Z-Space, which is specially optimized for TK-350 imagery. The particularity of DTM processing for TK-350 imagery consists in large amount of DTM points; it is usually equal to tens millions. The developed DTM generation method uses terrain and image pyramids to produce a dense disparity map with resolution up to one pixel. The matching speed achieves 1500 image points per second on ordinary computers such as IBM PC with Pentium 200MHz. The high matching speed is reached by prediction from a DTM pyramid with the help of a geomorphologic map. The regular DTM matrix is produced in geodesic northing -easting coordinates system by fast bilinear interpolation of 3D points. The speed of full DTM generation process achieves 700 DTM points per second. To produce very large DTM limited only by available hard drive space a procedure for merging of overlapped DTMs is developed. The accuracy of the DTM is measured in ground control points. The typical error variance is 10m in height.
INTERNATIONAL ARCHIVES OF …, Jan 1, 2000
In the paper we propose the method of object detection based on analysis of orthophoto images dif... more In the paper we propose the method of object detection based on analysis of orthophoto images difference. Object is registered stereoscopically on smooth surface of arbitrary shape. Orthophoto images for some basic plane are generated from left and right images of the stereopair. Orthophoto generation is based on a priori surface model knowledge. In case of object presence in the scene the orthophoto difference image contains a characteristic structure, which is detected by the matched filtration.
Proceedings Eurosip, Jan 1, 2007
Proceedings of SPIE Electronic Imaging …, Jan 1, 2007
The users of digital cameras often take multiple photographs of the same scene. Such multiple sho... more The users of digital cameras often take multiple photographs of the same scene. Such multiple shots usually have a special meaning to the photographer, and require further actions, e.g. selection of the best exposure/composition/portrait or stitching several images into a panorama or composite image. We present a method of fast retrieval of all groups of shots taken from the same viewpoint. This task is different from the recently emerged near-duplicate detection problem because, in our case, the multiple shots differ not only by photometric and simple geometric transformations; they can have a little or no overlap, and large variations of objects may be presented. Therefore, we solve a general multiple image registration problem by extracting local image descriptors, their matching, and recovering geometric transformation between images. Initially, the photo-collection is divided in time-based clusters, which are then refined by extracting connected components from the global image registration graph. The method has been applied to real consumer photo-collections, and we show that depending on individual camera usage styles, user collections contain from 15% to 90% of photos requiring further attention. The presented system automates the otherwise manual work of selecting a series of similar images.
SPIE Proc. vol. 3084, …, Jan 1, 1997
The technology for three-dimensional model design of real world scenes and its photorealistic ren... more The technology for three-dimensional model design of real world scenes and its photorealistic rendering are current topic of investigation. Development of such technology is very attractive to implement in vast variety of applications: military mission planning, crew training, civil engineering, architecture, virtual reality entertainments-just a few was mentioned. 3D photorealistic models of urban areas are often discussed now as upgrade from existing two-dimensional geographic information systems. Possibility of site model generation with small details depends on two main factors: available source dataset and computer power resources. In this paper PC based technology is presented, so the scenes of middle resolution (scale of 1:1000) be constructed. Types of datasets are the gray level aerial stereo pairs of photographs (scale of 1:14000) and true color on ground photographs of buildings (scale ca.1:1000). True color terrestrial photographs are also necessary for photorealistic rendering, that in high extent improves human perception of the scene.
Proceedings of SPIE, Jan 1, 2006
Virtual Storytelling. Using Virtual …, Jan 1, 2003
In this paper we present a new workflow allowing the creation of 3D characters in an automated wa... more In this paper we present a new workflow allowing the creation of 3D characters in an automated way that does not require the expertise of an animator. This workflow is based of the acquisition of real human data captured by 3D body scanners, which is them processed to generate firstly animatable body meshes, secondly skinned body meshes and finally textured 3D garments.
Uploads
Papers by Alexander Sibiryakov