Two-dimensional sine-assisted empirical mode image decomposition method

Abstract

The invention discloses a two-dimensional sine-assisted empirical mode image decomposition method. Fast empirical mode decomposition (EMD) is carried out on an image, two-dimensional sine distributions which are similar to a scale of a component is automatically designed through carrying out amplitude-frequency analysis on an initial decomposition result of the scale of a layer, an extremum distribution of the scale is enabled to be uniform through assistance of the sine distributions, and thus extremum can be completely and thoroughly separated out from the image. Repeated iteration operations are carried out, thus the image can be decomposed into a series of images of clear scales, explicit physical meanings and all components. Adopting the sine-assisted method can successfully solve themode mixing problem. In addition, the method requires only the pair of mutually inverse sine distributions to participate in operations, thus higher image decomposition efficiency can be guaranteed,and the method is enabled to have very good application potential. The method of the invention is high in self-adaptability, simple to realize and fast in a decomposition speed, and can be widely applied in the fields of image processing, pattern recognition, artificial intelligence and the like.

Description

technical field [0001] The invention relates to an image processing method, in particular to an image multi-scale decomposition method. Background technique [0002] In many technical fields that require image processing, such as machine vision and artificial intelligence, it is usually necessary to extract and analyze one or more components in an image, which usually uses multi-scale and multi-resolution technology in order to use different scales in the image to represent different image components. Commonly used image decomposition techniques include wavelet multi-resolution analysis, window Fourier transform method and so on. The biggest challenge facing the image decomposition method based on Fourier analysis is: when the image details are large or the content structure changes are complex, it is difficult to accurately analyze and extract the various spectral components of the image, while wavelet multi-resolution analysis needs to be determined according to the image...

AI Technical Summary

Problems solved by technology

However, this method will be disturbed by the problem of mode aliasing, so that the decomposed single-mode function is not composed of a single-scale signal but a variety of mixed-scale signals, which brings difficulties to the further application and analysis of the decomposition results; The mode aliasing problem is generally caused by the uneven distribution of extreme values ​​at a certain scale. In order to solve this problem, Ensemble Empirical Mode Decomposition (EEMD) adds a large amount of white noise, and then adds white noise each time After the noise, the single-mode results of EMD decomposition are added and averaged to eliminate the influence of the added white noise. This method is more robust, but the cost is to add as much white noise as possible, resulting in very poor timeliness , cannot be applied in practice

The mode aliasing problem of two-dimensional Empirical Mode Decomposition (BEMD) is similar to the one-dimensional case, but the timeliness problem of two-dimensional ensemble empirical mode decomposition (BEEMD) is even more difficult to meet the actual needs

To sum up, if the content of the image is complex or the image contains noise, the practical application of the existing image multi-scale decomposition technology will face a greater test

Images