Planar Nuclear Medicine Images De-Noising Via Wavelet Block Thresholding: a Simulation Study

Abstract

Aims: Wavelet transform is the powerful mathematical tool used for image processing and noise suppression applications in different area of science and engineering. In this technique, selecting optimal threshold for de-noising is still an area of thrust for the researchers. In this paper, we have focused on the de-noising of planar nuclear medicine image using the block thresholding with different block sizes when threshold values vary for Stein, soft and hard thresholdings. Study Design: De-noising of planar nuclear medicine images via wavelet block thresholding Place and Duration of Study: University of Guilan. Methodology: We simulated planar images of hot region of Carlson phantom by GATE v. 6.1. Noisy image and reference image were produced by imaging time 10 and 40 second, from the hot region of Carlson phantom which is placed next to the simulated gamma camera. Then, we tried to de-noise noisy test image by wavelet transforms and block thresholding methods. For de-noising, we show the evolution of the de-noising peak signal to noise ratio when threshold values vary for Stein, soft and hard thresholding methods. Results: We observed that for the given noisy image, the optimal thresholds belong to the soft and Stein thresholding algorithms, respectively. Comparing the different size of blocks for the soft block thresholder by the PSNR and RMSE criterions show that the best results could be obtained by test image which is subjected to the block sizes of 3 and 4. Furthermore, Invariant soft thresholding is found to yield an overly smoothed estimate than orthogonal soft thresholding. Conclusion: Although decreasing of imaging time increases Poisson noise in the acquired nuclear medicine images, using of de-noising technique based on the wavelet transform could improve image degradation, so that the quality of de-noised test image could be compared to the reference image.