Dedicated to machine vision software development
VISION fOr VISION specializes in the development of software for machine vision applications. We implement standard and custom vision algorithms, tune them for best performance, adapt them you your platforms and assemble complete solutions that suit your needs.
If you couldn't find an off-the-shelf package to address your vision problem, this is where VISION fOr VISION can help you by providing consultancy, doing research & development and writing efficient code for you.
We enjoy twenty years of active development of image processing algorithms and machine vision products. Our mission is to help you expand your business in connection to computer vision by providing taylor-made software.
Below are listed some of our recent achievements.
Aztec Code reader
The challenge: develop a custom detector/decoder for Aztec codes to be used from a handheld reader.
The benefit: allow reliable reading of large codes in difficult conditions such as partial occlusion.
Text detection on a mobile phone screen
The challenge: find a piece of text displayed on a screen with arbitrary location/size/rotation, before passing it to an OCR software.
The benefit: allow recognition regardless how the characters are presented to the camera, improving application robustness.
Edge movement measurement from a distance
The challenge: follow the real-time movement (4000 measurements per second) of a contrasted edge, with calibration, at a 20 m distance.
The benefit: perform accurate and fast non-contact measurement using standard vision components.
Fluorescence gauging in a test tube
The challenge: follow the evolution in time of fluorescent emission in a test tube.
The benefit: ease of experimentation with a fully automated process and good visual feedback.
Particle tracking in a reactor
The challenge: process large images (4 Mpixel @ 30 fps) in real-time in order to detect a tiny particle and follow its trajectory. SSE instructions were used to speed-up blob analysis.
The benefit: no need for custom hardware acceleration.
Image pre-warping for skewed projection
The challenge: warp an image in real-time to correct perspective deformation due to projector misalignement, with shading adjustments.
The benefit: make composite monumental projection possible with off-the-shelf hardware.
On the fly saw tooth inspection
The challenge: allow on-the-fly processing of the image from a linear scan camera in a way that is easily amenable to FPGA processing.
The benefit: cost reduction of the inspection device.
Barcode and Data Matrix reader for mobile phones
The challenge: make the reader portable to different mobile platforms, C++ and Java based, while working in uncontrolled viewing conditions – natural lighting and out of focus.
The benefit: our customer has complete control over the implementation details and features of the reader.
Real-time 3D registration of point clouds
The challenge: align several 3D point clouds in real-time, allowing for the six translation and rotation degrees of freedom, being robust to outliers, in real-time.
The benefits: by simplifying the registration problem and precomputing as much of what can be computed, the parallel structure of the algorithm could be enhanced to allow for a low-cost hardware implementation.
Fast correlation-based pattern matching
The challenge: accelerate the computation of the normalized grayscale correlation coefficient without missing the local maxima when subsampling.
The benefit: an extremely fast, MMX/SSE accelerated algorithm, running on a standard PC platform.
Car plate reading
The challenge: locate the plate(s) anywhere in the image with minimum a priori information and recognize the characters in difficult viewing conditions (poor lighting, low contrast, motion blur, interleaving…)
The benefits: a real-time solution that allows a standard PC to monitor several cameras with a sustained frame rate.