Suganthi S S, Ramakrishnan S
Indian Institute of Technology Madras.
Biomed Sci Instrum. 2014;50:328-35.
Analyses of breast thermograms are still a challenging task primarily due to the limitations such as low contrast, low signal to noise ratio and absence of clear edges. Therefore, always there is a requirement for preprocessing techniques before performing any quantitative analysis. In this work, a noise removal framework using fast non-local means algorithm, method noise and median filter was used to denoise breast thermograms. The images considered were subjected to Anscombe transformation to convert the distribution from Poisson to Gaussian. The pre-denoised image was obtained by subjecting the transformed image to fast non-local means filtering. The method noise which is the difference between the original and pre-denoised image was observed with the noise component merged in few structures and fine detail of the image. The image details presented in the method noise was extracted by smoothing the noise part using the median filter. The retrieved image part was added to the pre-denoised image to obtain the final denoised image. The performance of this technique was compared with that of Wiener and SUSAN filters. The results show that all the filters considered are able to remove the noise component. The performance of the proposed denoising framework is found to be good in preserving detail and removing noise. Further, the method noise is observed with negligible image details. Similarly, denoised image with no noise and smoothed edges are observed using Wiener filter and its method noise is contained with few structures and image details. The performance results of SUSAN filter is found to be blurred denoised image with little noise and also method noise with extensive structure and image details. Hence, it appears that the proposed denoising framework is able to preserve the edge information and generate clear image that could help in enhancing the diagnostic relevance of breast thermograms. In this paper, the introduction, objectives, materials and methods, results and discussion and conclusions are presented in detail.
乳房热成像分析仍然是一项具有挑战性的任务，主要是由于存在诸如对比度低、信噪比低和边缘不清晰等局限性。因此，在进行任何定量分析之前，始终需要预处理技术。在这项工作中，使用了一种基于快速非局部均值算法、方法噪声和中值滤波器的去噪框架来对乳房热成像进行去噪。所考虑的图像经过安斯库姆变换，将分布从泊松分布转换为高斯分布。通过对变换后的图像进行快速非局部均值滤波获得预去噪图像。观察到方法噪声，即原始图像和预去噪图像之间的差异，其噪声成分融入了图像的一些结构和精细细节中。通过使用中值滤波器平滑噪声部分，提取出方法噪声中呈现的图像细节。将检索到的图像部分添加到预去噪图像中，以获得最终的去噪图像。将该技术的性能与维纳滤波器和SUSAN滤波器的性能进行了比较。结果表明，所有考虑的滤波器都能够去除噪声成分。所提出的去噪框架在保留细节和去除噪声方面表现良好。此外，观察到方法噪声中的图像细节可以忽略不计。同样，使用维纳滤波器观察到无噪声且边缘平滑的去噪图像，其方法噪声包含的结构和图像细节较少。发现SUSAN滤波器的性能结果是去噪图像模糊且噪声较小，其方法噪声具有广泛的结构和图像细节。因此，所提出的去噪框架似乎能够保留边缘信息并生成清晰的图像，这有助于提高乳房热成像的诊断相关性。本文详细介绍了引言、目标、材料与方法、结果与讨论以及结论。
