Georgios Kordelas

Enhanced disparity estimation in stereo images

This paper presents a novel stereo disparity estimation method, which combines three different cost metrics, defined using RGB information, the CENSUS transform, as well as Scale-Invariant Feature Transform coefficients. The selected cost metrics are aggregated based on an adaptive weight approach, in order to calculate their corresponding cost volumes. The resulting cost volumes are then merged into a combined one, following a novel two-phase strategy, which is further refined by exploiting scanline optimization. A mean-shift segmentation-driven approach is exploited to deal with outliers in the disparity maps. Additionally, low-textured areas are handled using disparity histogram analysis, which allows for reliable disparity plane fitting on these areas. Finally, an efficient two-step approach is introduced to refine disparity discontinuities. Experiments performed on the four images of the Middlebury benchmark demonstrate the accuracy of this methodology, which currently ranks first among published methods. Moreover, this algorithm is tested on 27 additional Middlebury stereo pairs for evaluating thoroughly its performance. The extended comparison verifies the efficiency of this work. A two-phase strategy for combining separate cost volumes is described.A mean-shift segmentation-driven approach to handle disparity outliers is utilized.Low-textured area plane fitting is fostered by using disparity histogram analysis.Our method ranks first among published methods in the Middlebury evaluation system.

Reconstruction for 3D immersive virtual environments

The future of tele-conferencing is towards multi-party 3D Tele-Immersion (TI) and TI environments that can support realistic inter-personal communications and virtual interaction among participants. In this paper, we address two important issues, pertinent to TI environments. The paper focuses on techniques for the real-time, 3D reconstruction of moving humans from multiple Kinect devices. The off-line generation of real-life 3D scenes from visual data, captured by non-professional users is also addressed. Experimental results are provided that demonstrate the efficiency of the methods, along with an example of mixing real with virtual in a shared space.

Content-Based Guided Image Filtering, Weighted Semi-Global Optimization, and Efficient Disparity Refinement for Fast and Accurate Disparity Estimation

This paper presents a novel approach, which relies on content-based guided image filtering and weighted semi-global optimization for fast and accurate disparity estimation. The approach uses a pixel-based cost term that combines gradient, Gabor-Feature, and color information. The pixel-based matching costs are filtered by applying guided image filtering, which relies on rectangular support windows of two different sizes. In this way, two filtered costs are estimated for each pixel. Among the two filtered costs, the one that will be finally assigned to each pixel depends on the local image content around this pixel. The filtered cost volume is further refined by exploiting weighted semi-global optimization, which improves the disparity estimation accuracy. Finally, the disparity refinement in outlier regions relies on a straightforward and time-efficient outliers handling scheme and on a simple approach which deals with the disparity outliers at depth discontinuities. Experimental results on the Middlebury online stereo evaluation benchmark and 27 additional Middlebury stereo pairs prove that our method is able to generate disparity maps with high accuracy while keeping the computational cost low.

Viewpoint independent object recognition in cluttered scenes exploiting ray-triangle intersection and SIFT algorithms

Viewpoint independent recognition of free-form objects and estimation of their exact position are a complex procedure with applications in robotics, artificial intelligence, computer vision and many other scientific fields. In this paper a novel approach is presented that addresses recognition of objects lying in highly cluttered and occluded scenes. The proposed procedure relies on distance maps, which are extracted and stored off-line for each of the 3D objects that might be contained in the scene. During the on-line recognition procedure distance maps are extracted from the scene. Greyscale images, derived from scenes distance maps, are matched with those of the object under recognition by applying similarity measures to the descriptors that are extracted from the images. The similarity is then estimated from image patches, which are defined using the SIFT descriptor in an appropriate way. After finding the best similarities the position of the object in the scene is estimated. This process is repeated until all objects are successfully recognized. Multiple experiments, which were performed on both 2.5D synthetic and real scenes, proved that the proposed method is robust and highly efficient to a satisfactory degree of occlusion and clutter.

Robust SIFT-based feature matching using Kendall's rank correlation measure

The Scale Invariant Feature Transform, SIFT, is one of the most efficient image matching techniques based on local features. It has been applied to various scientific domains such as machine vision, robot navigation, object recognition, etc. In this work, a SIFT improvement is proposed that makes feature matching more robust in the presence of different types of image noise. Thus, Kendalls rank correlation measure is employed to improve the performance of feature matching. Its exploitation reduces the number of erroneous SIFT feature matches without adding significantly to the execution time. The results of the SIFT improvement are validated through matching examples between similar images.

Recognizing 3D objects in cluttered scenes using projection images

This paper presents a novel descriptor for recognizing objects in highly occluded and cluttered 2.5D scenes produced by range scans. This new compact regional shape descriptor, called “projection images”, is designed to be robust against noise, partial occlusion and clutter. Projection images are formed by “projections” of points onto the plane centered at the basis point which is perpendicular to the viewing axis. Multiple experiments were performed on a dataset of 50 range scans, each one comprised of multiple objects, which proved that the proposed method is robust and efficient to a satisfactory degree of occlusion and clutter, while it compared favorably against descriptors previously introduced in the literature.

Revisiting guided image filter based stereo matching and scanline optimization for improved disparity estimation

In this paper the scanline optimization used for stereo matching, is revisited. In order to improve the performance of this semi-global technique, a new criterion to check depth discontinuity, is introduced. This criterion is defined according to the mean-shift-based image segmentation result. Additionally, this work proposes the employment of a pixel dissimilarity metric for the computation of the cost term, which is then provided to the guided image filter approach to estimate the initial cost volume. The algorithm is tested on the four images of the online Middlebury stereo evaluation benchmark. Moreover, it is tested on 27 additional Middlebury stereo pairs for assessing thoroughly its performance. The extended comparison verifies the efficiency of this work.

Recognizing 3D Objects using Ray-Triangle Intersection Distances

A novel method for recognizing 3D objects in an occluded, cluttered and noisy 2.5D scene, is presented. A ray-triangle intersection algorithm is used to compute distances between a circular sector that does not belong to the object and a triangulated surface. Firstly, for each sectors point its distance from the object is calculated and stored in a distance map. Secondly, a 2D histogram that counts the distance maps points whose corresponding distance falls within its distance bins, is formed. Then, the percentages of the bin points that fall within each bin are calculated forming the final descriptor vector. The same procedure is followed for the 2.5D scene. The number of the extracted descriptor vectors is independent to the number of the objects or scenes vertices. Experiments proved that the proposed method is fast, robust to noise, occlusion and clutter.

Accurate stereo 3D point cloud generation suitable for multi-view stereo reconstruction

This paper proposes a novel methodology for generating 3D point clouds of good accuracy from stereo pairs. Initially, the methodology defines some conditions for the proper selection of image pairs. Then, the selected stereo images are used to estimate dense correspondences using the Daisy descriptor. An efficient two-phase strategy to remove outliers is then introduced. Finally, the 3D point cloud is refined by combining sub-pixel accuracy correspondences estimation and the moving least squares algorithm. The proposed methodology can be exploited by multi-view stereo algorithms due to its good accuracy and its fast computation.