Sparse-representation-based denoising of photoacoustic images

Abstract

Optical resolution photoacoustic microscopy (OR-PAM) is an emerging hybrid technology that combines optical contrast and acoustic resolution. The quality of photoacoustic (PA) images is degraded due to different parameters such as frequency, the diameter of the transducer or external noise induced from the laser. The diameter of the transducer is proportional to its near field to focus or unfocus the transducer, which affects the image quality, so reconstruction and denoising of photoacoustic images is an important issue in medical imaging, especially when it comes to diagnosing diseases at an early stage. Visualization of different structures in the PA images requires filtering to suppress noise. This paper investigates the use of K-means singular value decomposition (K-SVD) to eliminate noise and enhance the effect of vasculature in the PA images. The algorithm is tested on PA images of blood-filled tubes of different diameters and in vivo mouse ear images acquired using OR-PAM imaging. The results reveal a better denoising capability of PA images when compared with standard Wiener and wavelet-based filtering.