Determination of the Optimum Threshold Value in a Denoising Method with a Wavelet Transform

摘要

The reduction of dose in an x-ray image reduces the patient's risk, but at the same time, the x-ray image is degraded with Poisson noise. To reduce the Poisson noise in the x-ray image, we proposed a new method with wavelet shrinkage in 2009. This method uses scaling coefficients of a wavelet transformed image and estimates the noise variance pixel by pixel. This method works well for various kinds of images; however we have to optimize the threshold value depending on the statistical data of the pixel values when we want to get a more effective result. In this paper we proposed a determination method for the optimum threshold value from the transformed coefficients. To evaluate the value determined by the proposed algorithm, we acquired several x-ray images with different doses and investigated the relationship between the peak signal to noise ratio (PSNR) and determined threshold value. The results of experiments showed that the determined optimum value did not always agree with the threshold value that yields the largest PSNR, but the quality of details in the image improved using the determined value with keeping the performance of noise reduction.