Self-organizing features for regularized image standardization.

摘要

Image standardization is an important preprocessing step in several image processing applications. In neuroimaging, by reducing normal variability through the standardization of brains, functional activity from multiple subjects can be overlaid to study localization. Furthermore, variability outside normal ranges can be used to report abnormalities. In automatic facial expression recognition, by standardizing the facial features, the accuracy of the facial expression recognition can be increased. The current standardization methods are mostly based on global alignment and warping strategies. However, global standardization methods fail to align individual structures accurately.; In this study, we propose a feature-based, semi-automatic, non-parametric, and non-linear standardization framework to complement the existing global methods. The method consists of three phases: In phase one, templates are generated from the atlas structures, using Self-Organizing Maps (SOMs). The parameters of each SOM are determined using a new topology evaluation technique. In phase two, the atlas templates are reconfigured using points from individual features, to establish a one-to-one correspondence between the atlas and individual structures. During training, a regularization procedure can be optionally invoked to guarantee smoothness in areas where the discrepancy between the atlas and individual feature is high. In the final phase, difference vectors are generated using the corresponding points of the atlas and individual structure. The whole image is warped by interpolation of the difference vectors through Gaussian radial basis functions, which are determined by minimizing the membrane energy.; Results are demonstrated on simulated features, as well as selected sulci in brain MRIs, and facial features. There are two significant advantages of this system over the existing standardization schemes: increased accuracy and speed in the alignment of internal features. Although our framework does not handle standardization of global shape and size differences, it can easily be used as a complementary module for the existing global standardization techniques, to increase precision of local alignment.