Computer Vision & Image Processing Consulting

We specialize in consulting services for challenging problems in computer vision and image analysis. We believe that good deep learning tools start with good feature engineering and properly pre-processed data. We use our unique approach to image processing based on curve- and orientation-fields. Our company has worked with challenging data in a wide areas of applications, including biological, medical and seismic imaging, and have published several papers in the field.

Some of our consulting work includes:

  • Developing real time object detection algorithms for movie data.
  • Developing text enhancement algorithms for data with wildly varying contrast and flash glare artifacts.
  • Removing grid artifacts from fiber optics footage.
  • Developing denoising algorithms for ultra sound medical imaging.

Our algorithms for using orientation and curve fields are based on the paradigm to describe lower level features geometrically and topologically, as opposed based on intensity. Although there are many examples of shape-based image descriptors in the literature, our tools for generating orientation- and curve-fields are based on a novel integration based method, which makes our methods extremely robust with respect to low signal-to-noise ration, poor resolution, and non-Gaussian artifacts such as compression artifacts. Our method can also be used to segment objects in scenes with rapid contrast and lighting changes in the scene.

Example of segmentation of a cryo electron microscope image using generalized orientation fields. From The Generalized Orientation Field Transform in Object Modeling, Algorithms, and Applications, Barneva, R.P. et al Ed., Research Publishing (2010)

Denoising example using curve fields. From The Curve Filter Transform: A Robust Method for Curve Enhancement Sandberg, K., Lecture Notes in Computer Science (LNCS), 6454, p. 107-116 (2010)For movie illustrating the Curve Filter Transform (CFT, a.k.a. the Curve Field Transform).

For links to some of our publications, see the list of our company publications.

For an example of our unique curve field based denoising method, see this movie.