Аналiз якостi методiв повороту зображення зa допомогою моментних iнварiантiв
Вантажиться...
Файли
Дата
2020
Автори
Бедратюк, Г.І.
Bedratyuk, A.
Назва журналу
Номер ISSN
Назва тому
Видавець
Хмельницький національний університет
Анотація
У роботі запропоновано новий метод оцінки якості повороту напівтонових зображень на основі статистичних
поворотних інваріантів зображення. Поворотні моменти довільного порядку не змінюються при повороті зображень і тому
можуть слугувати мірою для оцінки зміни якості зображень при поворотах. Проведені числові експерименти показали, що
поворотні моменти ефективно виявляють спотворення зображення і поводять себе аналогічно, як і відомі індекси якості
зображень.
Digital image processing often results in a loss of image quality, which is expressed in the appearance of unwanted distortion on the converted image. Visual image quality measurement is important for image and video processing programs. There are about a hundred different metrics, measuring image quality, the most popular of which are root mean square error (MSE), peak signal to noise ratio (PSNR), universal image quality index (UQI), structural similarity index (SSIM), descriptor index similarity (FSIM), gradient similarity measure (GSM), noise level measure (NQM). One of the most common standard image processing methods is to rotate an image centered at any point in the plane. In fact, rotation is a resampling of an image followed by interpolation. There have been many different rotation algorithms recently, which makes it important to study methods for assessing their quality. In this article, we propose a new method for estimating the quality of rotation based on geometric invariant moments of the image. Geometric moments and their invariants were first introduced and are intensively used in computer vision and pattern recognition. Global image descriptors are built on moments, which characterize the shape of objects in the image and are resistant to simple transformations – parallel transfer, rotation and uniform scaling. In this paper, we offer a set of invariants that are resistant only to turns. The fact that the torques do not change for the original and perfectly interpolated image allows you to use the torques as a measure of the quality of interpolation when rotating. The paper proposes a new method for estimating the quality of rotation of grayscale images based on statistical rotational invariants of the image. The geometric moment invariants of arbitrary order do not change under rotating images and therefore can serve as a measure to assess the image quality. Numerical experiments have shown that the invariants effectively detect image distortion and behave similarly to known image quality indices.
Digital image processing often results in a loss of image quality, which is expressed in the appearance of unwanted distortion on the converted image. Visual image quality measurement is important for image and video processing programs. There are about a hundred different metrics, measuring image quality, the most popular of which are root mean square error (MSE), peak signal to noise ratio (PSNR), universal image quality index (UQI), structural similarity index (SSIM), descriptor index similarity (FSIM), gradient similarity measure (GSM), noise level measure (NQM). One of the most common standard image processing methods is to rotate an image centered at any point in the plane. In fact, rotation is a resampling of an image followed by interpolation. There have been many different rotation algorithms recently, which makes it important to study methods for assessing their quality. In this article, we propose a new method for estimating the quality of rotation based on geometric invariant moments of the image. Geometric moments and their invariants were first introduced and are intensively used in computer vision and pattern recognition. Global image descriptors are built on moments, which characterize the shape of objects in the image and are resistant to simple transformations – parallel transfer, rotation and uniform scaling. In this paper, we offer a set of invariants that are resistant only to turns. The fact that the torques do not change for the original and perfectly interpolated image allows you to use the torques as a measure of the quality of interpolation when rotating. The paper proposes a new method for estimating the quality of rotation of grayscale images based on statistical rotational invariants of the image. The geometric moment invariants of arbitrary order do not change under rotating images and therefore can serve as a measure to assess the image quality. Numerical experiments have shown that the invariants effectively detect image distortion and behave similarly to known image quality indices.
Опис
Ключові слова
якість зображення, поворот, інтерполяція зображення, індекси якості, поворотні інваріанти, image quality, rotation, image interpolation, quality indices, rotation invariants
Бібліографічний опис
Бедратюк Г. І. Аналiз якостi методiв повороту зображення зa допомогою моментних iнварiантiв / Г. І. Бедратюк // Вимірювальна та обчислювальна техніка в технологічних процесах. – 2020. – № 2. – С. 57-64.