Feature based recognition of photonic devices in images obtained by confocal scanning laser microscopy

In microscopy the detection, recognition and description of objects contained in images constitute important problems. Recently there have been developed a series of techniques for invariant feature detection/description. These techniques can provide the detection of a set of key points in one image, key points that are invariant to effects like rotation, scale changes, noise, illumination conditions and even view-point. Several comparisons have been conducted for natural images in order to find out which of the techniques provides best performance. In this paper we describe a method for the recognition of images representing photonic quantum ring lasers obtained by scanning laser microscopy. Our feature-based method relies on the scale invariant feature transform technique (SIFT), which at this time is considered to be one of the most robust methods in the detection of invariant features. The recognition of devices represents an important step for developing a registration framework that will consist of several stages : invariant feature detection (with SIFT), feature matching, transform model estimation and image alignment. Once the images are registered, comparisons over the image content will be easier to be conducted, faster, and more reliable.