| Image have more and more influence on people’s lives, and it is an important means for human beings to communicate. However, images will be polluted inevitably by the noise what affect the visual effect of image recognition in the collection and transmission, because the performance of electronic components is not enough. So image denoising is a very important part of image processing. The primary purpose of image denoising is to preserve the structure information of the image, and then to reduce the image noise. Whaterer the spatial domain denoising algorithm or the frequency domain denoising algorithm, often ignore the image edge details, resulting in the image becomes blurred.The block-matching and 3D filtering algorithm is one of the most excellent algorithm,which solves the problem of image noise reduction. However due to the hard thresholding for image matching of the similar of neighborhood block, in the higher intensity of image noise,the threshold is relatively low, that lead to the similar image blocks number is not enough,than image denoising effect is not ideal; under lower intensity of image noise, threshold for higher image similarity block matching number, image noise reduction time too long.After studying the basis and theory of image denoising, and aiming at the shortcomings of BM3 D noise reduction, this paper proposes an improved algorithm, which is based on the adaptive threshold of BM3 D. The core of the algorithm is using the adaptive threshold distance instead of the hard threshold in the original BM3 D method in the image process what obtain the similar neighborhood blocks. The standard image is divied into blocks, and pointwise calculation of gradient-based structural similarity and Eulidean distance of BM3 D is performed, and the threshold of matching is the Euclidean distance of the most similarimage blocks. The image denoising using this algorithm has achieved good results, whether it is objective evaluation or subjective visual observation. |
PDF Full Text Request