Patrice Brault

A set of template matching predictors for intra video coding

Prediction methods by template matching are often mentioned to improve video coding efficiency. They are based on a Markovian model to find the most similar patterns of texture in previously encoded information. These kinds of methods are more efficient than H.264/AVC intra prediction modes in many cases, such as complex texture coding. However, the template matching method is not always optimal as to the best predictor choice. This paper introduces the use of a sorted set of template matching predictors to propose different texture patterns. At the encoder side, the best predictor is selected by minimizing a rate-distortion metric and is transmitted to the decoder. Despite this needed additional side information, simulation results show that the proposed scheme further improves the coding efficiency of H.264/AVC up to 11%.

Advances in measuring ocean salinity with an optical sensor

Absolute salinity measurement of seawater has become a key issue in thermodynamic models of the oceans. One of the most direct ways is to measure the seawater refractive index which is related to density and can therefore be related to the absolute salinity. Recent advances in high resolution position sensitive devices enable us to take advantage of small beam deviation measurements using refractometers. This paper assesses the advantages of such technology with respect to the current state-of-the-art technology. In particular, we present the resolution dependence on refractive index variations and derive the limits of such a solution for designing seawater sensors well suited for coastal and deep-sea applications. Particular attention has been paid to investigate the impact of environmental parameters, such as temperature and pressure, on an optical sensor, and ways to mitigate or compensate them have been suggested here. The sensor has been successfully tested in a pressure tank and in open oceans 2000 m deep.

Gaussian mixture model-based contrast enhancement

In this paper, a method for enhancing low contrast images is proposed. This method, called Gaussian Mixture Model based Contrast Enhancement (GMMCE), brings into play the Gaussian mixture modeling of histograms to model the content of the images. Based on the fact that each homogeneous area in natural images has a Gaussian-shaped histogram, it decomposes the narrow histogram of low contrast images into a set of scaled and shifted Gaussians. The individual histograms are then stretched by increasing their variance parameters, and are diffused on the entire histogram by scattering their mean parameters, to build a broad version of the histogram. The number of Gaussians as well as their parameters are optimized to set up a GMM with lowest approximation error and highest similarity to the original histogram. Compared to the existing histogram-based methods, the experimental results show that the quality of GMMCE enhanced pictures are mostly consistent and outperform other benchmark methods. Additionally, the computational complexity analysis show that GMMCE is a low complexity method.

Automated transformation-invariant shape recognition through wavelet multiresolution

We present here new results in Wavelet Multi-Resolution Analysis (W-MRA) applied to shape recognition in automatic vehicle driving applications. Different types of shapes have to be recognized in this framework. They pertain to most of the objects entering the sensors field of a car. These objects can be road signs, lane separation lines, moving or static obstacles, other automotive vehicles, or visual beacons. The recognition process must be invariant to global, affine or not, transformations which are : rotation, translation and scaling. It also has to be invariant to more local, elastic, deformations like the perspective (in particular with wide angle camera lenses), and also like deformations due to environmental conditions (weather : rain, mist, light reverberation) or optical and electrical signal noises. To demonstrate our method, an initial shape, with a known contour, is compared to the same contour altered by rotation, translation, scaling and perspective. The curvature computed for each contour point is used as a main criterion in the shape matching process. The original part of this work is to use wavelet descriptors, generated with a fast orthonormal W-MRA, rather than Fourier descriptors, in order to provide a multi-resolution description of the contour to be analyzed. In such way, the intrinsic spatial localization property of wavelet descriptors can be used and the recognition process can be speeded up. The most important part of this work is to demonstrate the potential performance of Wavelet-MRA in this application of shape recognition.

Bayesian Wavelet Domain Segmentation

We have recently demonstrated that fully unsupervised segmentations of still images and 2D+T sequences is possible by Bayesian methods, on the basis of a Hidden Markovian Model (HMM) and a Potts‐Markov Random Field (PMRF), in the pixel domain. The use of a high number of iterations to reach convergence in a segmentation where the number of segments, or “classes” labels, is important makes the algorithm rather slow for the processing of a large quantity of data like in image sequences. We more recently have worked out a new version of this algorithm by first operating our segmentation in the wavelet transform domain rather than in the direct domain. Doing so, we take advantage of the local decay property, or “peaky” distribution of the wavelet coefficients, in an orthogonal decomposition. This decomposition is a fast pyramidal. O(N2), decomposition, so the Bayesian segmentation is performed only once on the first coarse image then on all sub‐bands up to the highest resolution level. Moreover, we have impro...

A new construction for motion and speed capture with conical wavelets

Motion analysis and in particular, speed and rotation analysis, has been introduced in the 80s using the continuous wavelet transform(CWT) with Morlet wavelets. The motion-tuned WT appeared to be an efficient framework and an alternative to the optical flow (OF), the block matching (BM) or the phase difference, for the study of motion. In particular it has shown better performances in the case of noise, in the case of occlusions and in long temporal-dependency motions. However, the Morlet wavelet has defects that cause some difficulties when performing an analysis. By construction, Cauchy and conical wavelets alleviate these drawbacks. In particular, they can be oriented without interference on other parameters and characteristics of the filter. We have extended these wavelets to the spatio-temporal domain and to motion parameters, and we have studied the capabilities of this new Gaussian-Conical-Morlet wavelet (GCM). We present here the first analysis results with this new construction.