Noise suppressing method capable of keeping multi-spectral image thin-wire structure

Abstract

The invention discloses a noise suppressing method capable of keeping a multi-spectral image thin-wire structure. The noise suppressing method includes the following steps of a, robust analysis of the partial direction, wherein under the condition of pulse noise interference, the partial direction of the edge and the thin-wire structure in a multi-spectral image is obtained; b, building of a partial weighting template, wherein the weights of all pixels in a neighborhood are determined according to the partial direction; c, direction weighted vector median filtering, wherein the weighting distance is calculated according to the weights, the weighting distance sum is calculated firstly, then the pixels are ranked according to the weighting distance sum, and the pixel value with the smallest weighting distance sum is found and serves out output. Pulse noise can be suppressed, the thin-wire structure in the image can be effectively kept, and the noise suppressing method is particularly suitable for noise reduction and characteristic strengthening of images and data with multiple channels like common colorful natural scene images, multi-spectral remote sensing images, medical multi-spectral images and seismic exploration data volumes.

Description

technical field [0001] The present invention relates to a multispectral image noise reduction method, in particular to a noise suppression method capable of maintaining the thin line structure of the multispectral image, which is suitable for such as ordinary color natural scene images, multispectral remote sensing images, medical multispectral images, earthquake Noise reduction and feature enhancement for images and data with multiple channels, such as exploration volumes. Background technique [0002] In the process of acquisition and transmission of images and scientific data, interference will inevitably be encountered, such as sensor noise, background interference noise, transmission channel interference, etc. These interferences are generally referred to as noise. There are many forms of noise, Gaussian noise and impulse interference noise are the most common. For Gaussian noise, the traditional linear filter has a better suppression effect. Impulse interference nois...

AI Technical Summary

Problems solved by technology

However, the existing filtering methods do not consider the local structure of the image, so that the vector median filter cannot be adaptively adjusted according to the direction of the image itself.

This defect removes some of the fine line structure of the image

Images