Heinz Hügli

Computing visual attention from scene depth

Visual attention is the ability to rapidly detect the interesting parts of a given scene. Inspired by biological vision, the principle of visual attention is used with a similar goal in computer vision. Several previous works deal with the computation of visual attention from images provided by standard video cameras, but little attention has been devoted so far to scene depth as source for visual attention. The investigation presented in this paper aims at an extension of the visual attention model to the scene depth component. The first part of the paper is devoted to the integration of depth in the computational model built around conspicuity and saliency maps. The second part is devoted to experimental work in which results of visual attention, obtained form the extended model and for various 3D scenes, are presented. The results speak for the usefulness of the enhanced computational model.

A multi-resolution ICP with heuristic closest point search for fast and robust 3D registration of range images

The iterative closest point (ICP) algorithm is widely used for the registration of 3D geometric data. One of the main drawbacks of the algorithm is its quadratic time complexity O(N/sup 2/) with the number of points N. Consequently, several methods have been proposed to accelerate the process. We present a new solution for the speeding up of the ICP algorithm and special care is taken to avoid any tradeoff with the quality of the registration. The proposed solution combines a coarse to fine multiresolution approach with the neighbor search algorithm. The multiresolution approach permits to successively improve the registration using finer levels of representation and the neighbor search algorithm speeds up the closest point search by using a heuristic approach. Both multiresolution scheme and neighbor search algorithm main features are presented. Confirming the success of the proposed solution, typical results show that this combination permits to create a very fast ICP algorithm, with a closest point search complexity of O(N), while preserving the matching quality.

Assessing the contribution of color in visual attention

Visual attention is the ability of a vision system, be it biological or artificial, to rapidly detect potentially relevant parts of a visual scene, on which higher level vision tasks, such as object recognition, can focus. The saliency-based model of visual attention represents one of tile main attempts to simulate this visual mechanism on computers. Though biologically inspired, this model has only been partially assessed in comparison with human behavior. Our methodology consists in comparing the computational saliency map with human eye movement patterns. This paper presents an in-depth analysis of the model by assessing the contribution of different cues to visual attention. It reports the results of a quantitative comparison of human visual attention derived from fixation patterns with visual attention as modeled by different versions of the computer model. More specifically, a one-cue gray-level model is compared to a two-cues color model. The experiments conducted with over 40 images of different nature and involving 20 human subjects assess the quantitative contribution of chromatic features in visual attention.

Usefulness of the LPC-residue in text-independent speaker verification

Abstract This paper is a contribution to automatic speaker recognition. It considers speech analysis by linear prediction and investigates the recognition contribution of its two main resulting components, namely the synthesis filter on one hand and the residue on the other hand. This investigation is motivated by the orthogonality property and the physiological significance of these two components, which suggest the possibility of an improvement over current speaker recognition approaches based on nothing but the usual synthesis filter features. Specifically, we propose a new representation of the residue and we analyse its corresponding recognition performance by issuing experiments in the context of text-independent speaker verification. Experiments involving both known and new methods allow us to compare the recognition performance of the two components. First we consider separate methods, then we combine them. Each method is tested on the same database and according to the same methodology, with strictly disjoint training and test data sets. The results show the usefulness of the residue when used alone, even if it proves to be less efficient than the synthesis filter. However, when both are combined, the residue shows its true relevance. It achieves a reduction of the error rate which, in our case, went down from 5.7% to 4.0%.

A multi-resolution scheme ICP algorithm for fast shape registration

The iterative closest point (ICP) algorithm is widely used for the registration of geometric data. One of its main drawbacks is its quadratic time complexity O(N/sup 2/) with the shape number of points N, which implies long processing time, especially when using high resolution data. This paper proposes to accelerate the process by a coarse to fine multiresolution approach in which a solution at a coarse level is successively improved at a finer level of representation. Specifically, it investigates this multiresolution ICP approach when coupled with the tree search or the neighbor search closest point algorithms. A theoretical and practical analysis and a comparison of the considered algorithms are presented. Confirming the success of the multiresolution scheme, the results also show that this combination permits us to create a very fast ICP algorithm, gaining speed up to a factor of 27 over a standard fast ICP algorithm.

Multi-feature matching algorithm for free-form 3D surface registration

Applications such as object digitizing, object recognition and object inspection need efficient surface matching algorithms. Several variants of an iterative closest point (ICP) matching algorithm have been proposed for such tasks. This paper proposes and analyzes a multi-feature ICP matching algorithm that includes the surface color and the surface orientation information. The matching error minimization keeps the original closed-form solution. Therefore, the convergence of the multi-feature ICP algorithm cannot be proven any more. However, experiments show successful convergence. Further experimental results applying the multi-feature ICP to free-form objects show a significant increase of the range of successful convergence range.

Adaptive color image compression based on visual attention

This paper reports an adaptive still color image compression method which produces automatically selected ROI with a higher reconstruction quality with respect to the rest of the input image. The ROI are generated on-the fly with a purely data-driven technique based on visual attention. Inspired from biological vision, the multicue visual attention algorithm detects the most visually salient regions of an image. Thus, when operating in systems with low bit rate constraints, the adaptive coding scheme favors the allocation of a higher number of bits to those image regions that are more conspicuous to the human visual system. The compressed image files produced by this adaptive method are fully compatible with the JPEG standard, which favors their widespread utilization.

Geometric matching of 3D objects: assessing the range of successful initial configurations

This paper considers the matching of 3D objects by a geometric approach based on the iterative closest point algorithm (ICP), which, starting from an initial configuration of two rigid objects, iteratively finds their best correspondence. The algorithm does not converge always to the best solution. It can be trapped in a local minimum and miss the optimum matching. While the convergence of this algorithm towards the global minimum is known to depend largely on the initial configuration of test and model objects, this paper investigates the quantitative nature of this dependence. Considering the space C of relative configurations of the two objects to be compared, we call range of successful initial configurations, or SIC-range, the subspace of C which configurations bring the algorithm to converge to the global minimum. In this paper, we present a frame for analyzing the SIC-range of 3D objects and present a number of original experimental results assessing the SIC-range of a number of real 3D objects.

Fast ICP Algorithms for Shape Registration

Shape registration plays an important role in applications such as 3D object modeling or object recognition. The iterative closest point (ICP) algorithm is widely used for the registration of geometric data. One of its main drawback is its time complexity O(N2), quadratic with the shape size N, which implies long processing time, especially when using high resolution data. Several methods were proposed to accelerate the process. One of the most effective one uses a tree search (k-D tree) to establish closest point relationships and reduces the complexity to O(N logN). This paper reviews several of the existing methods and proposes and analyses a new, even less complex ICP algorithm, that uses a heuristic approach to find the closest points. Based on a local search it permits to reduce the complexity to O(N) and to greatly accelerate the process. A comprehensive analysis and a comparison of the considered algorithm with a tree search method are presented.

Generation And Use Of Color Pseudo Random Sequences For Coding Structured Light In Active Ranging

When using structured light for the measurement of three dimensional objects, the observed stripes of light must be indexed, that is, the correspondance between projected light stripes and the observed stripes must be established. Basically, time, spatial and color coding can be used to characterize the light stripes. In this study, we use color coding and generate the projected light stripes as a sequence of colors showing subsequences of length N which are all different and thus can be indexed by analysis of the observed stripe sequence. The sequence generation used follows the methods known for generating pseudo random sequences. We present also several indexing rules and analyse their behaviour in a measurement setup exhibiting stripe deletions and permutations. A parametric method is proposed which can accomodate scenes with different object coarseness.