Fundamentally, objects within digitised images have a dimension that is greater then one so each point of the object has a position in space (and possibly time) that is different from all other points and an associated direction. In general the possible directions will correspond to a circle, sphere or hypersphere of the appropriate dimension.

In practice, the accurate estimation of orientation is a key issue in the support of a wide range of applications of image analysis such as in the differentiation of a shape from a background, the recognition and tracking of shapes and importantly (for example in the case of monitoring or treatment of tumours) the quantification of growth or shrinkage of the object.

THE OXFORD INVENTION

At present there are a number of difficulties associated with the estimation of object orientations in an image, with image noise being the most challenging. The main advantage of this new technique is the ability to work at multiple spatial scales, which is a requirement for successful use in many applications. The Oxford team has been able to overcome the problems associated with noise by using a combination of steerable filters and accurate orientation computation. They have also demonstrated that the technique is applicable:

-- Image matching

-- Stereo vision

-- Object tracking

Another important application for the technology is detecting locally rounded objects, which enables, for example, the identification of the presence of calcifications in 2D mammograms (an indication that breast cancer may be present) and well-defined structures such as larger tumours in 3D and 4D medical datasets.
 

http://www.gzjiayumed.com/en/index.asp