Frequency domain weighting correlation method for image registration

Abstract

The invention discloses a frequency domain weighting correlation method for image registration, which comprises the following steps of: windowing two images to be registered to acquire images for inhibiting frequency spectra from leaking; performing discrete Fourier transform on the images to acquire the frequency spectra and removing a high-frequency part from the frequency spectra of the images by using a low-pass filter; substituting the remaining frequency spectra into a phase weighting correlation expression to obtain a correlation peak frequency domain expression, performing inverse discrete Fourier transform to obtain the correlation peak of a pixel domain and finding coordinates with a maximum value to obtain the displacement value of the whole pixel; and performing binary quartic surface fitting on the periphery of a peak value, performing interpolation calculation to obtain the displacement value of a sub-pixel and adding the displacement value of the whole pixel and the displacement value of the sub-pixel to obtain the registration displacement of sub-pixel accuracy between the two images. The frequency domain weighting correlation method for the image registration has high registration accuracy on digital images and high robustness on scenes and signal-to-noise ratio, and is suitable for the field of the processing of various digital images.

Description

technical field [0001] The invention relates to the technical field of computer image registration, in particular to a related method for frequency domain weighting for image registration. Background technique [0002] Obtaining the relative displacement between two images is an important research content of image registration, which is widely used in various digital image processing fields, such as image stitching, super-resolution reconstruction of remote sensing images, high dynamic contrast (HDR) image construction, real-time Image stabilization technology, motion blur image restoration, etc. [0003] With the development of digital image processing technology, many image registration methods have emerged. In summary, these methods can be roughly divided into: direct methods (Irani M, Anandan P.About Direct Methods.Lecture notes in computer science.Springer.1999: 267-277) and feature-based methods (Torr P H S, Zisserman A. Feature based methods for structure and motion ...

AI Technical Summary

Problems solved by technology

[0007] However, the above derivation is based on ideal conditions and does not take into account the influence of noise and non-ideal sampling on the method

For real images, due to frequency aliasing and spectral leakage caused by noise and non-ideal sampling, the accuracy of traditional phase correlation algorithms will be greatly affected

In particular, non-ideal sampling has a huge impact on accuracy, and sometimes even makes the algorithm completely invalid

The traditional phase correlation algorithm has the following defects: (1) When calculating the phase difference of two images in the frequency domain, the weights of different frequencies are the same, without considering the energy distribution in the frequency domain. For general natural images, the spectral energy Mainly distributed in the low frequency, it is shown in the image that the low frequency represents most of the information; in (1), due to the existence of the normalized denominator, the phase difference weights calculated at different frequency points are the same, that is, regardless of the high Frequency component or low frequency component, their phase difference has the same impact on the final result, which obviously deviates from the distribution law of image content

Moreover, frequency components with small amplitudes are more susceptible to noise and frequency aliasing, which is not conducive to the robustness of the algorithm

(2) The algorithm does not deal with the possible frequency mixing, and this error is finally included in the relevant results

Because in actual imaging, the scene is generally not an ideal frequency-limited signal, so it does not meet the conditions of Nyquist sampling, and the image has more or less frequency aliasing problems

Especially for cameras with larger photosensitive device pixel size, the phenomenon of aliasing is more serious due to its low spatial sampling rate

(3) As an image algorithm based on frequency domain processing, it does not take into account the frequency domain spectrum leakage caused by rectangular truncation in the pixel domain

However, the actual signal processing system cannot achieve a large sampling frequency and cannot process a lot of data.

In addition, many signals themselves may contain frequency components of the full frequency band, and it is impossible to increase the sampling frequency to infinity

Images