earticle

영어

The main problem faced during biomedical image diagnosis is the noise introduced due to the consequence of the coherent nature of the image. The noise interfered may be Gaussian noise, speckle noise or Poisson noise, during transmission. The capturing devices itself has a salt & pepper noise. These noises corrupt the image and often lead to incorrect diagnosis. These noises make it more difficult for the observer to discriminate fine detail of the images in diagnostic examinations. Thus, denoising these noises from a noisy image has become the most important step in medical image processing. In spite of the sophistication of the recently proposed methods, most algorithms have not yet attained a desirable level of applicability. Denoising techniques are aimed at removing noise or distortion from images while retaining the original quality of the image. In this work, we propose PCA_NLM approach which computes neighborhood similarities after PCA projection. Our algorithm is based on the assumption that image contains an extensive amount of self-similarity. The accuracy and computational cost of the PCA algorithm is improved by computing neighborhood similarities, i.e., averaging weights, after a PCA projection to a lower dimensional subspace. We evaluate and compare the performance of proposed technique with different existing methods by using six quality measures PSNR, SNR, MSE, NAE, Correlation Coefficient and SSIM. Comparative analysis shows our approach give the best performance results in terms of improved quality measures as well as visual interpretation.