Search

In this paper we present a method for interactive high-quality visualization of large and time-varying 3D unstructured data from particle simulations.

Recent work on multi-object tracking has shown the promise of tracklet-based methods. In this work we present a method which infers tracklets then groups them into tracks. It overcomes some of the disadvantages of existing methods, such as the use of heuristics or non-realistic constraints. The main idea is to formulate the data association problem as inference in a set of Bayesian networks. This avoids exhaustive evaluation of data association hypotheses, provides a confidence estimate of the solution, and handles split-merge observations. Consistency of motion and appearance is the driving force behind finding the MAP data association estimate. The computed tracklets are then used in a complete multi-object tracking algorithm, which is evaluated on a vehicle tracking task in an aerial surveillance context. Very good performance is achieved on challenging video sequences. Track fragmentation is nearly non-existent, and false alarm rates are low.

Video from aerial surveillance can provide a rich source of data for many applications and can be enhanced for display and analysis through such methods as mosaic construction, super-resolution, and mover detection. All of these methods require accurate frame-to-frame registration, which for live use must be performed in real time. In many situations, scene parallax may make alignment using global transformations impossible or error-prone, limiting the performance of subsequent processing and applications. For these cases, dense (per-pixel) correspondence is required, but this can be computationally prohibitive. This paper presents a hierarchical dense correspondence algorithm designed for implementation on graphics processing units (GPUs). Since the method does not rely on epipolar geometry, it is also suitable for use when there are uncor-rected nonlinear lens distortions. A framework for using this dense correspondence to implement local mosaicking, super-resolution enhancement, and mover detection is also presented and demonstrated using examples of each of these types of enhancement and different types of video sources.

Summary form only given. We developed the piecewise-linear Haar (PLHaar) transform (Senecal, J.G. et al., Proc. 12th Pacific Conf. on Computer Graphics and Appl., p.371-80, 2004), an integer wavelet-like transform. PLHaar does not have dynamic range expansion, i.e., it is an n-bit to n-bit transform. To our knowledge, PLHaar is the only reversible n-bit to n-bit transform that is suitable for lossy and lossless coding. We are investigating PLHaar's use in lossy image coding. Preliminary results from thresholding transform coefficients show that PLHaar does not produce objectionable artifacts like previous n-bit to n-bit transforms, such as that of Honyang Chao et al. (CFH) (see Advances in Computational Mathematics, Lect. Notes In Pure and App. Math., vol.202, p.13-38, 1999). Also, at lower bitrates PLHaar images have increased contrast. For a given set of CFH and PLHaar coefficients with equal entropy, the PLHaar reconstruction is more appealing, although the PSNR may be lower.

We introduce the piecewise-linear Haar (PLHaar) transform, a reversible n-bit to n-bit transform that is based on the Haar wavelet transform. PLHaar is continuous, while all current n-bit to n-bit methods are not, and is therefore uniquely usable with both lossy and lossless methods (e.g. image compression). PLHaar has both integer and continuous (i.e. non-discrete) forms. By keeping the coefficients to n bits PLHaar is particularly suited for use in hardware environments where channel width is limited, such as digital video channels and graphics hardware.

Arithmetic coding achieves a superior coding rate when encoding a binary source, but its lack of speed makes it an inferior choice when true high-performance encoding is needed. This paper presents the practical implementation of fast entropy coders for binary messages utilizing only bit shifts and table lookups. To limit code table size the proposed code lengths is limited with a type of variable-to-variable (VV) length code created from source string merging. This is referred to as "merged codes". With merged codes it is possible to achieve a desired level of speed by adjusting the number of bits read from the source at each step. The most efficient merged codes yield a coder with a worst-case inefficiency of 0.4%, relative to the Shannon entropy. Using a hybrid Golomb-VV bin coder the compression ratio that is competitive with other state-of-the-art coders, at a superior throughput is achieved.

Methods of producing reversible n-bit to n-bit integer transforms is presented in this paper. Such methods are particularly suited for hardware-based implementations, as keeping the coefficients to n bits simplifies the design of custom hardware and makes it easier to use these approaches on preexisting hardware with limited channel width. One of our methods, called table-lookup Haar (TLHaar) is an approximation of the Haar integer wavelet transform (Haar IWT). The Haar IWT takes two integer data values A and B and using averaging and differencing produces a low-pass value L and a high-pass value H, both integers. The lookup tables are created by initializing each with an identity transform. TLHaar executes up to 44% faster. Compression results are mixed, and depend on the compression method used and the image type.