Poisson noise reduction in scintigraphic images using Gradient Adaptive Trimmed Mean filter

Abstract

We propose a new hybrid technique for reduction of poisson noise in scintigraphic images. Our proposed method is a combination of Gradient calculation and Adaptive Trimmed Mean filter (ATMF). In a predefined window, gradient of the center pixel is averaged out. ATMF remove the lowest and highest variations in the pixel values of Gradient denoised image and average out remaining neighborhood pixel values. The proposed technique is applied on scintigraphic images. Results are compared with conventional filters i.e. Median, Wiener filter and latest denoising filter i.e. Non Local Mean (NLM) filter. The proposed scheme shows good visual results with improving Correlation, Mean Squared Error (MSE), Structural Similarity Index Metric (SSIM) and Peak to Signal Noise Ratio (PSNR) of the image.