Christophe Montagne

Global interior robot localisation by a colour content image retrieval system

We propose a new global localisation approach to determine a coarse position of a mobile robot in structured indoor space using colour-based image retrieval techniques. We use an original method of colour quantisation based on the bakers transformation to extract a two-dimensional colour pallet combining as well space and vicinity-related information as colourimetric aspect of the original image. We conceive several retrieving approaches bringing to a specific similarity measure integrating the space organisation of colours in the pallet. The bakers transformation provides a quantisation of the image into a space where colours that are nearby in the original space are also nearby in the output space, thereby providing dimensionality reduction and invariance to minor changes in the image. Whereas the distance provides for partial invariance to translation, sight point small changes, and scale factor. In addition to this study, we developed a hierarchical search module based on the logic classification of images following rooms. This hierarchical module reduces the searching indoor space and ensures an improvement of our system performances. Results are then compared with those brought by colour histograms provided with several similarity measures. In this paper, we focus on colour-based features to describe indoor images. A finalised system must obviously integrate other type of signature like shape and texture.

Adaptive filtering and hypothesis testing: Application to cancerous cells detection

We propose a new method to detect cells in microscopic imagery, the problem under study being the analysis of cancerous cells experiencing metastasis i.e. cells susceptible to migration. This work would help medical researchers to study the evolution of a cancer. The peculiar nature of the images due to the acquisition protocol causes some difficulties. These are resolved through tailored preprocessing involving correction of uneven illumination and enhancement of cellular information. Detection and counting of cells are performed by our proposed filtering that provokes peaks in its convolution space wherever cells are present. We compare our counting results with those provided by human experts and with a Hough transform developed for similar purposes. The validity of the cell segmentation from the peaks is then established by a statistical test of the closeness of the segmented cell to a cell model.

Detection and Counting of "in vivo" cells to predict cell migratory potential

In this paper, we present a work which is performed by biologists and computer scientists both. The aim of this work is to evaluate the migratory potential of cancerous cells. Cancer is characterised by primary tumour. When some cells move they create new tumours, which are called metastases. It is very important to understand this migration process in order to be able to arrest it and increase the chances of a cure. Today, biologists analyse images from different cell cultures and manually count one by one the cells present therein. It is a hard and fastidious work, so here we present some algorithms to automatically perform these tasks of detection and counting. The images that we have are very low contrasted, with a gradient of illumination, and the cells are numerous and tightly aggregated. In this paper different algorithms are evocated and results compared for about 150 images comprising more than 65,000 cells.

Adaptive dynamic time warping for recognition of natural gestures

Gesture recognition is one of the important tasks for human System Interaction (HRI). This paper describes a novel approach intended to recognize 3D dynamic composed gestures by combining Dynamic Time Warping (DTW) with an Adaptive Sliding Window which the name Adaptive Dynamic Time Warping (ADTW). We use the skeleton algorithm provided by the Kinect SDK to track the upper part of body and extract joints angles based on depth information. Each gesture is represented by the combination of angles variations and stored described as a vector. A composed gesture is a sequence of two simple gestures or more performed successively in time. We chose five simple gestures : come, recede, point to the right, point to the left and stop. For each simple gesture, we chose a reference sequence that perfectly represents it. In order to recognize all gesture of the composed gesture in the right order, we combine (DTW) with an Adaptive Sliding Window. In one hand, we use an adaptive window to browse through the sequence of the composed gesture by feeding it to each time with new data. In other hand, we use DTW to compare between the reference gestures and the the sequences defined by the adaptive window. In fact, by comparing each two sequences, DTW computes the euclidean distance between them. Finally, the reference gesture which gives the lower distance is considered as the source class of the tested gesture.

Quality dependent multimodal fusion of face and iris biometrics

Although iris is known as the most accurate and face as the most accepted in biometrics, these distinct modalities encounter variability in data in real-world applications. Such limitation can be overcome by a multimodal system based on both traits. Additionally, by conditioning the multimodal fusion on quality, useful information can be extracted from lower quality measures rather than rejecting them out of hand. This paper suggests a dynamic weighted sum fusion that exploits an iris occlusion-based quality metric while combining unimodal scores. Instead of incorporating the quality of the gallery and probe images separately, a single quality metric for each gallery-probe comparison was used. Two strategies for integrating this metric into score-level fusion were explored. Experiments on the IV2 multimodal database including multiple variabilities proved that the proposed method improves some best current non quality-based fusion schemes by more than 30% in terms of Equal Error Rates.

Combination of HMM and DTW for 3D Dynamic Gesture Recognition Using Depth Only

Gesture recognition is one of the important tasks for human Robot Interaction (HRI). This paper describes a novel system intended to recognize 3D dynamic gestures based on depth information provided by Kinect sensor. The proposed system utilizes tracking for the upper body part and combines the hidden Markov models (HMM) and dynamic time warping (DTW) to avoid gestures misclassification. By using the skeleton algorithm provided by the Kinect SDK, body is tracked and joints information are extracted. Each gesture is characterized by one of the angles which remains active when executing it. The variations of the angles throughout the gesture are used as inputs of Hidden Markov Models (HMM) in order to recognize the dynamic gestures. By feeding the output of (HMM) back to (DTW), we achieved good classification performances without any misallocation. Besides that, using depth information only makes our method robust against environmental conditions such as illumination changes and scene complexity.

Bio-inspired image enhancement derived from a ‘rank order coding’ model

In this study, the authors propose a new method to enhance image information, based on wavelet decomposition and original partial reconstruction of image. This reconstruction called ‘asynchronous reconstruction’ is not carried out in the same way as the usual sequential one. It is based on rank order coding. In fact, while sequential reconstruction is to sum all or a part of the responses obtained for each scale of ‘coarse to fine’ decomposition, asynchronous reconstruction tries to be closer to human brain which uses a limited number of frequency channels. Actually, after wavelet decomposition, responses are sorted from top down for each pixel of the image. Final asynchronous reconstruction for each pixel is obtained by adding a chosen number of wavelet responses, beginning by the maximum response. So, at a given level of reconstruction, the pixel values do not come from the same frequency channels. The interest of this method has been tested on a face verification task using the IV2 biometric database. Stopping criterion for reconstruction can be a constant number of wavelet responses to use, but an adaptive process has been also investigated. Three criteria are explored: standard deviation, entropy and lost edges ratio.

Suitability of Digital Elevation Models for Watershed Segmenting Images with Directional Illumination

This paper investigates the use of different functions for the digital elevation model input to the watershed transform. The use of gradient information is the most frequent one, but its strength varies due to illumination variations. We investigate the two major classes of input functions, distance maps and the gradient, their combinations, and propose an different function using soft clustering memberships that is not covariant with illumination.