Projects

“SmartCaption” is a software that aims to minimize human labor during the captioning of a video, in either "roll-up" or "pop-on" mode. The software enables the execution of three main tasks of a subtitling session: text editing, text synchronization to facilitate its reading and automatic positioning to prevent display on key visual elements (text or human face). This project was funded in part by NSERC.

Since 4 years, the team participates on a regular basis to the TREC Video Retrieval Evaluation Technology of the National Institute of Standards and Technology (NIST). This "competition" is an opportunity for groups of international researchers to compare the performance of their video processing technologies on a common datasets. The Vision and Imaging team has had the opportunity to polish his video analysis tools for security applications in video surveillance and developed, in collaboration with the CRIM’s Speech Recognition team, technologies to detect fraudulent video copy.

Since 2007, the team has served several mandates to develop technologies for using video in training elite athletes. This type of application requires many various expertises, including the installation of multi-camera capture system and the development of software for handling, analysis and simulation of video data. These mandates are being made for users of Speed ​​Skating Canada and the Centre National Multisport de Montréal. 

This project's objective is to evaluate an automated tree crown detection library and obtain performance improvements by migrating certain components to free software, thereby providing the bases for production software development.

These mandates, made ​​for Natural Resources Canada, have led to the development of a software tool for the semi-automatic update of the contour maps of the Canadian hydrographic network (rivers and lakes) using SPOT satellite imagery . A web version running in service mode is also under development.

The purpose of this mandate is the implementation of software to process and manipulate large volumes of digital microfilm of textual material (newspapers, magazines, brochures, etc.).

The software combines various functions such as filtering, enhancement, cropping and rectification that are required for archiving and indexing applications. 

The goal of the E-Inclusion Research Network is to develop audiovisual processing tools and provide guidelines so content producers can deepen the multimedia experiences of people with auditory or visual impairments. The project is financed in part by Heritage Canada.

Within two E-Inclusion projects, the team contributes to the development of tools used in the intelligent positioning of subtitles for the hard-of-hearing, as well as those used to provide descriptive video of visual content for the visually impaired.

• "Des sous-titres mieux adaptés". Stéphane Gagné, Découvrir, mai-juin 2007. (PDF 226 KB)  
• "Des images qui en disent long". Sybille Pluvinage, Découvrir, décembre 2007-janvier 2008. (PDF 124 KB)

The goal of the COBVIS project is to develop a simulation environment for the analysis of cephalo-ocular behaviours in elderly drivers. It should enable the creation of improved rehabilitation programs and develop the tools required to better identify at-risk drivers. The SPEED-Q project aims at adapting this simulator to an automobile platform. The project leader is the Computer Vision and Systems Laboratory (CVSL) at Université Laval.

The team contributes to the project through the automatic detection and recognition of facial expressions under difficult driving conditions using the “Facial Action Coding System”. The project is financed by AUTO21 Network of Centres of Excellence and Société d'Assurance Automobile du Québec (SAAQ).

The goal of this project is to simulate artificial polarimetric radar (PolSAR) images to reproduce the expected statistical features as accurately as possible. This project was financed by the NSERC over a five-year period.

Subsequently, new techniques used for multi-scale analysis in geometric spaces are currently being implemented and will be used to provide new tools to information from PolSAR images.

The goal of the PTZ project was to develop an intelligent Pan/Tilt/Zoom camera for video surveillance applications. The project leader was VideoWave Network.

The CRIM team partnered in the development and implementation of tracking and object recognition algorithms. PTZ was financed in part by the Precarn-CRIM Alliance program.

The project received the 2006 OCTAS Award for Technological Innovation, jointly presented to VideoWave Networks and CRIM by the Fédération de l'Informatique du Québec, now Réseau Action TI (an organisation representing members from the Quebec IT industry).

• Video - demonstration (AVI 10.9 MB)

The purpose of the SIMIPE-Ciné project was to develop landscape recognition software designed to assist in filming location searches.

The team was involved in developing the content image search engine. The project leader was the Chair in Landscape and Environmental Design (Université de Montréal). Financing was provided by the CRD Montérégie (Conseil régional de développement de la Montérégie).

The goal of the C³GRID project was to develop a computing grid for the distributed learning of acoustic, visual and speech recognition language models.

The team contributed to the visual features extraction module (based on shapes of the mouth) in order to improve recognition performance in a foley sound effects environment. This project was financed in part by the CANARIE-ARIM program.

The MONNET project was developed by the Computer Vision and Systems Laboratory of Université Laval. The goal was to develop an intelligent vision system for the surveillance of people in public places.

The CRIM team contributed to the development of the face and facial expression recognition module. MONNET project was financed in part by the Precarn University-Led (or PUL) Program.

• "Mon ordinateur sait quand je souris". France Laliberté et Langis Gagnon, @ccéléraTIon, automne 2005. (PDF 252 KB) 

The goal of the S3I project was to develop a visual inspection station for small plastic objects. The project leader was the I.C. Vision Company of Montréal. To fine-tune the computer image acquisition and processing module, the team partnered with the Département de génie de la production automatisé of École de technologie supérieure.

This project was financed in part by the Precarn-CRIM Alliance program.

• Video presentation of the project (MPG 47 MB)
• "De la science derrière le bouchon". Louise Desautels, Découvrir, jan.-fév. 2004. (PDF 69 KB)

The objective of the MADIS project was to develop a test bench for the indexation and content search of films based on the MPEG-7 standard.

For this project, the National Film Board of Canada (NFB) partnered with the CRIM Vision and Imaging and Speech Recognition teams. MADIS was financed in part by the CANARIE E-content program

• Video presentation of the project (MPG 99.2 MB)
•   "À la recherche des sons et des images". Louise Desautels, Découvrir, nov.-déc. 2003. (PDF 104 KB)
  

The goal of this research project was to explore audio and video combination approaches used to recognise instances of Québec French spoken during newscasts. The project was also aimed at developing a test environment for the extraction of audio-visual descriptors.

It was conducted in collaboration with the CRIM’s Speech Recognition team.

This project produced a system that automatically creates MPEG-7 descriptions allowing for the indexation of image banks based on visual features (colour, form, texture, etc.).

The search, analysis and description browsing capabilities and image classification functions, were also developed.

• "MPEG-7: un nouveau standard pour le multimédia". Valérie Gouaillier, @ccéléraTIon, Winter 2003. (PDF 60 KB)
  

The team developed algorithms to process and analyze ophthalmic images in this collaborative effort headed by Université de Montréal.

It was partner in an infrastructure subsidy with the Canada Foundation for Innovation (CFI).

The purpose of this mandate was to develop an application for automatic or semi-automatic segmentation of lesions appearing in fluorescent imaging sequences and provide an estimate of their surface area.

• "Traitement d'image et infertilité". Louise Desautels, Découvrir, Mar.-April. 2003. (PDF 64 KB)
  

This contract consisted in developing a motor drive for the radiometric calibration of colour infrared and optical aerial photographs in the creation of image mosaics.

The mosaic development process included radiometric and colorimetric compensation operations, as well as operations relating to the layout and assembly of pre-processed component images (tiles).

The objective of this project was to build a toolbox of algorithms for colour ophthalmic images analysis used to assist in the diagnosis of retina disorders.

The main deliverable was a prototype software (Retsoft) used by medical specialists to assist diagnosing diabetic retinopathy.