Planar Scintigraphic Images Denoising

Scintigraphic images are generally affected by a Poisson type random noise which diminishes qualitatively and quantitatively the images. Restoration techniques aim to “find” an object from one (or several) degraded observation(s). The objective of the restoration is then to produce an image closer to the physical reality. So that the restoration is successful, it is very useful to know the nature of degradation. In this work, we present a planar scintigraphic acquisition chain modeling. This model takes into account the Poisson noise and its stationarity aspect. Then, we present a comparative study of the multi-resolution methods used to reduce the noise in scintigraphic images. Scintigraphy is a tool for exploring functionally several pathologies: the ventricular ejection fraction, the renal clearance and the thyroid activity. Given the fact that scintigraphic images are strongly affected by noise, the objective in this work is to enhance scintigraphic images for a reliable diagnosis and better orientation and understanding of the pathological phenomenon. This paper focuses on two main parts: the first deals with the degradation of model while the second takes into consideration the comparison of the multi-resolution methods for assessing the quality of scintigraphic images to reduce noise using wavelet, contourlet, curvelet, ridgelet and bandelet transformations.