SIAM IS14 - MS33 - Models and Methods for Imaging through Turbulence - Supplementary material

The focus of this minisymposium is the general problem of seeing through turbulent media of nonuniform density; for example the warm atmosphere or the oscillating surface of water. This is a wide field of research separated into a few sub-communities: astronomical seeing, underwater imaging, and long-distance surveillance over a hot terrain. Each of these problems has different challenges and constraints, but still there are many ideas that could be shared between them. The objective of this minisymposium is to bring together people of each of these backgrounds and discuss the differences and similarities between each turbulence model.

• 10:30-11:00 Atomic Models of Video Turbulence (online demo)
  We recall some simple methods to simulate turbulent videos of static objects. Each method is motivated by a physical model, but we describe them as image processing operators acting on the latent image. The latent image can be recovered easily from each simulation. Thus, we claim that the main difficulty in the correction of turbulence must come from a combination of different effects. Indeed, we find several of these individual effects appearing in real examples.
  Enric Meinhardt-Llopis, ENS Cachan, FRANCE
  Tristan Dagobert, Direction Generale de l'Armement, FRANCE
  
  
  
• 10:30-11:00 Independent Components in Dynamic Refraction
  Refraction causes random dynamic distortions in atmospheric turbulence and in views across a water interface. The latter scenario is experienced by submerged animals seeking to detect prey or avoid predators, which may be airborne or on land. Man encounters this when surveying a scene by a submarine or divers while wishing to avoid the use of an attention-drawing periscope. The problem of inverting random refracted dynamic distortions is difficult, particularly when some of the objects in the field of view are moving. On the other hand, in many cases, just those moving objects are of interest, as they reveal animal, human, or machine activity. Furthermore, detecting and tracking these objects does not necessitate handling the difficult task of complete recovery of the scene. We show that moving objects can be detected very simply, with low false-positive rates, even when the distortions are very strong and dominate the object motion. Localizing objects in three dimensions (3D) despite this random distortion is also important to some predators and also to submariners avoiding the salient use of periscopes. Refracted distortion statistics induce a probabilistic relation between any pixel location and a line of the salient use of periscopes. Refracted distortion statistics induce a probabilistic relation between any pixel location and a line of sight in space. Measurements of an objects random projection from multiple views and times lead to a likelihood function of the objects 3D location. The likelihood leads to estimates of the 3D location and its uncertainty. sight in space. Measurements of an objects random projection from multiple views and times lead to a likelihood function of the objects 3D location. The likelihood leads to estimates of the 3D location and its uncertainty.
  Marina Alterman Technion, Israel
  Yoav Schechner, Technion, Israel
  Pietro Perona, Caltech, USA
  Joseph Shamir, Technion, Israel
  
  
  
• 11:00 - 11:30 Video Restoration of Turbulence Distortion
  When the video is taken from a long range system, atmospheric turbulence can corrupt the video sequence and an object can look distorted. Blurring and diffeomorphism are couple of the main effects of atmospheric turbulence. We propose methods to stabilize the video sequence and give a good reference image. We reconstruct a new video sequence using Sobolev gradient sharpening with temporal smoothing, nd one latent image is found further utilizing the lucky-region method.
  Yifei Lou University of California, Irvine, USA
  
  
• 11:30 - 12:00 A Geometric Method for Image Recovery Through Optical Turbulence
  The phenomenon that is commonly referred to as optical “turbulence” in imaging is caused by the time and space-varying refraction index of the air which is due, among other factors, to temperature, air pressure, humidity, and wind conditions between the acquired scene and the image-capturing device. The resulting image sequence is also affected by the different and changing lighting conditions within the scene, by the actual distance between the observed objects and the camera, and by other artifacts introduced by the device itself. The above described distortion may be modeled, at least to a first approximation, as the combined effect of (i) a blur with an anisoplanatic point spread function and (ii) a time-dependent deformation of the image domain. In this talk I will describe an algorithm that, starting from this observation, first employs a geometric method for restoring the structure of the scene, and then uses variational deconvolution techniques to yields a crisp, final result. The algorithm may be viewed as an alternate minimization procedure of a functional that includes a data matching term, a regularization term for the deformations, and a regularization term for the recovered image. The algorithm has proven very effective for the the recovery of images affected by both ground-level atmospheric blur, and by underwater turbulence caused by temperature gradients.
  Mario Micheli Department of Mathematics, University of Washington, USA