In this paper, a model of the beam propagation is developed according to the physical properties of THz waves used in THz computed tomography (CT) scan imaging. This model is first included in an acquisition simulator to observe and estimate the impact of the Gaussian beam intensity profile on the projection sets. Second, the model is introduced in several inversion methods as a convolution filter to perform efficient tomographic reconstructions of simulated and real acquired objects. Results obtained with three reconstruction methods (BFP, SART and OSEM) are compared to the techniques proposed in this paper. We will demonstrate an increase of the overall quality and accuracy of the 3D reconstructions.

This paper addresses an application that has not been much explored, the de-identification of faces with expressions preservation in images. With the huge amount of images and videos shared on the Internet, protecting the identity of people in the data becomes crucial. Removing the identity information is often referred as de-identification. In this paper, we propose a novel de-identification process that preserves the important clues on the face for further behavior or emotions analysis. It is a difficult problem because obtaining the anonymity implies deteriorating the main face components. At the opposite, analyzing the expressions requires keeping enough information on the face such as, for instance, the gaze or the corners of the lips. Our approach relies on face and key points detection, followed by a variational adaptive filtering. Experimental results show the potential of the proposed method and open new insights for image dissemination or video broadcasting.

The main contribution of this paper is a new method for classifying document images by combining textual features extracted with the Bag of Words (BoW) technique and visual features extracted with the Bag of Visual Words (BoVW) technique. The BoVW is widely used within the computer vision community for scene classification or object recognition but few applications for the classification of entire document images have been submitted. While previous attempts have been showing disappointing results by combining visual and textual features with the Borda-count technique, we're proposing here a combination through learning approach. Experiments conducted on a 1925 document image industrial database reveal that this fusion scheme significantly improves the classification performances. Our concluding contribution deals with the choosing and tuning of the BoW and/or BoVW techniques in an industrial context.