July 02 Reviewed: August 2022
Download 0.5 Mb. Pdf ko'rish
|
50 13 34 1-6
METHODOLOGY
An important feature of the image noise attenuation method with minimal distortion of the image itself is that smoothing is carried out in the direction for which the brightness gradient is minimal, i.e. in the direction of light boundaries and lines [9; 181-182.]. For this purpose, the width of the pulse characteristic of the smoothing filter is chosen equal to one pixel, in while its length is taken quite large. As a result, the number of samples forming the impulse response also turns out to be large, which provides a significant filtering effect. Since when implementing this filtering method, the impulse response is always located in the direction of the minimum brightness change in the image, its length can be chosen large enough without fear of blurring the light boundaries, because it will not cross them, but will be located parallel to them. When implementing the described method, a set of pulse characteristics of various forms of the length specified in the dialog is generated before filtering begins. During the filtering process, the image, as usual, is scanned line by line. At the same time, for each new position of the central element of the impulse response in the image, its optimal shape is found. This is done by selecting a suitable instance from the set of its generated instances [4; 15-16]. The criterion for choosing the optimal instance is the minimum variance of the brightness of pixels overlapped by the pulse characteristic, i.e. from the entire set of pulse characteristics, the characteristic for which this variance is minimal is selected. The calculation of the variance for the case when all samples of the impulse response are the same and equal is performed 1/N according to the formula. ( ) 1 2 2 2 1 2 1 , N c их c N k L k L N − − =− = − Download 0.5 Mb. Do'stlaringiz bilan baham: |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling