School of Information Science and Technology, Nantong University, Nantong 226019, China.
School of Electrical and Automation Engineering, Nanjing Normal University, Nanjing 210046, China.
Sensors (Basel). 2023 Jul 6;23(13):6207. doi: 10.3390/s23136207.
Extracting the profiles of images is critical because it can bring simplified description and draw special attention to particular areas in the images. In our work, we designed two filters via the exponential and hypotenuse functions for profile extraction. Their ability to extract the profiles from the images obtained from weak-light conditions, fluorescence microscopes, transmission electron microscopes, and near-infrared cameras is proven. Moreover, they can be used to extract the nesting structures in the images. Furthermore, their performance in extracting images degraded by Gaussian noise is evaluated. We used Gaussian white noise with a mean value of 0.9 to create very noisy images. These filters are effective for extracting the edge morphology in the noisy images. For the purpose of a comparative study, we used several well-known filters to process these noisy images, including the filter based on Gabor wavelet, the filter based on the watershed algorithm, and the matched filter, the performances of which in profile extraction are either comparable or not effective when dealing with extensively noisy images. Our filters have shown the potential for use in the field of pattern recognition and object tracking.
提取图像轮廓至关重要，因为它可以带来简化的描述，并特别关注图像中的特定区域。在我们的工作中，我们设计了两个通过指数和斜边函数的滤波器来进行轮廓提取。事实证明，它们能够从弱光条件、荧光显微镜、透射电子显微镜和近红外相机获得的图像中提取轮廓。此外，它们还可以用于提取图像中的嵌套结构。此外，还评估了它们在提取高斯噪声退化图像方面的性能。我们使用均值为 0.9 的高斯白噪声来创建非常嘈杂的图像。这些滤波器可有效地提取噪声图像中的边缘形态。为了进行比较研究，我们使用了几种知名的滤波器来处理这些噪声图像，包括基于 Gabor 小波的滤波器、基于分水岭算法的滤波器和匹配滤波器，它们在轮廓提取方面的性能在处理广泛存在噪声的图像时要么相当，要么无效。我们的滤波器已经显示出在模式识别和目标跟踪领域应用的潜力。
