Computer Vision Techniques
With our image registration techniques, we can accurately align two or more images of an object, compensating for distortions caused by differences in viewing angle, distance, and orientation; sensor resolution; and shifts in object positions.
We can also 'stitch' multiple images with overlapping fields of view to produce a segmented panorama or high-resolution image.
Stitched Image: Retinal Panorama
We have wide-ranging experience in video processing, including compression, image stabilization, the detection of objects, video analytics, motion compensation and video tracking.
Our algorithms can detect and follow a subject through video footage, tracking the figure when occluded, identifying the figure surrounded by similar-looking subjects, and tracking the figure across footage from multiple cameras.
This technique is also used to detect defects in a certain substance or material, such as silicon chips.
Image by mattbuck. Licensed under CC.
Our Automatic Optical Inspection techniques are used in the semiconductor field to detect defects, and for the inspection of printed circuit boards (PCB) and flat panel display (FPD) surveillance.
Further, our nondestructive testing techniques provide a means of detecting and examining a variety of surface flaws, such as corrosion, contamination, surface finish, and surface discontinuities on joints, bonds and cracks.
Our pattern recognition techniques are used to determine patterns and regularities in images or video footage. Our software assigns input values to a given set of classes (for example, to determine whether a car is yellow or blue, or a post comment is "spam" or "not spam"). Many of our pattern recognition algorithms are probabilistic in nature, employing statistical inference to find the best label for a given instance.
Our character recognition software can learn a range of alphabets, and handwriting styles. We can take on complex optical character recognition (OCR) projects that regular OCR software fails to deal with accurately. We frequently undertake projects involving digitizing floating text and symbols in art works, for example.
We can accentuate or sharpen image features such as contrast or boundaries to make a graphic display more productive for display & analysis. This includes gray level & contrast manipulation, super resolution, noise reduction, edge sharpening, filtering, interpolation and magnification, pseudo-coloring and so on, as necessary to make images more useful for our clients.
Our 3D modeling techniques enable the user to view the object in any perspective from plane projections of the 3D volume. We can segment a part of the model for further analysis, segmenting the lower respiratory tract from the pulmonary system for closer examination, for instance.
Our 3D reconstruction software can replicate the 3D structure of the object and its surface with minimal information. From this, it can conduct measurements and calibrations, such as the width of an artery.
Our software can count objects even when only partially shown, such as the number of individuals in an image of a dense crowd. Our algorithms are designed to first ensure accurate detection, then count each detected category, for example the number of lymphocytes there are in a white blood cell sample.
Our machine learning software enables automatic prediction, classification, decision-making and diagnosis. Our work includes deep learning and convolutional neural networks (CNNs).
For a project example, see Machine Learning - Video Classification
Using our version of the Viola-Jones algorithm, we can create software that detects and recognizes faces, measuring distance between the eyes and the bridge of the nose, for example. Such software is useful for security systems, for patient identification, and personalized marketing.
Image: Wikimania2009 Beatrice Murch
CC BY 2.0 (http://creativecommons.org/licenses/by/2.0) via Wikimedia Commons
We develop a wide range of medical technologies including algorithms for catheter detection, cell classification, X-ray, CT, MRI, and ultrasound work, antigen detection, high-throughput testing, 3D reconstruction of organs, lung segmentation, and blood vessel quantitative measurements.
Sometimes camera quality falls short of what we need. RSIP Vision's software can act to calibrate your camera to give better imaging results. For example, we can correct lens distortion, or calibrate color distortion.