Office of Technology Transfer – University of Michigan

Fast Model-Free Super-Resolution Using a Two-Sensor Camera

Technology #4806

Questions about this technology? Ask a Technology Manager

Download Printable PDF

Categories
Researchers
Andrew Yagle
Managed By
Drew Bennett
Associate Director - Software Licensing 734-615-4004

UM File # 4806

Background
For good quality high resolution images, the digital sensor of an imaging device must be either large or highly sensitive to accommodate more pixels, but either option significantly increases the sensor’s price. Super-resolution (SR) is a computational method concerned with overcoming the limited resolution of the digital sensor of an imaging device. Motion-based SR methods are the most popular and widely used to model and solve this numerically large size, ill posed problem and they are in general computationally expensive. Motionless SR methods are also model-based and sensitive to model errors. A computationally efficient super-resolution method that can enhance the resolution of digital images, without the estimation of an underlying model, is highly desirable. Images with resolvable fine details are very important in many applications which make digital image enhancement a promising area of research for markets with total revenues > $2.
Technology Description
Researchers at the University of Michigan came up with a SR technique specifically designed for an imaging system equipped with two sensors with different pixel densities. The proposed method computes a super-resolved image from a plurality of low resolution images, without estimating the model of the underlying physical process that produced the low resolution images. Only the solution of a few small linear systems of equations, where the number of unknowns is equal to the number of low resolution images, is required, which translates to significant reduction in the computational cost. Proven by numerical experiments, this computationally cost-effective method completely bypasses the problems related to inaccurate model estimation, expanding its utility far beyond what is allowed by conventional super-resolution techniques.
Applications • Surveillance applications • Remote sensing • Ground-based planetary imaging • Satellite imaging • Medical imaging • Scientific photography
Advantages • Cheaper and simpler hardware requirement • Computationally cost-effective • Robustness to model errors • Wider applicability compared to conventional methods