Stochastic resonance aided robust techniques for segmentation of medical ultrasound images

The existence of stochastic resonance has been demonstrated in physical, biological and geological systems for boosting weak signals to make them detectable. Narrow regions, small features and low-contrast or subtle edges, in noisy images, correspond to such weak signals. In this paper, the occurrence and exploitation of stochastic resonance in the detection, extraction and analysis of such features is demonstrated both mathematically and empirically. The mathematical results are confirmed by simulation studies. Finally, results on medical ultrasound images demonstrate that several subtle features lost by the application of robust techniques such as mean shift filter are recovered by stochastic resonance. These results reconfirm the mathematical and simulation findings.