Texture analysis based on log polar frequency spectrum and Gabor;

In this paper,we incorporate Fourier transform with Log-Polar transform to achieve rotation invariant texture analysis.

This method can graph-display the measure result utilized the characteristic of the Log-Polar which is non-linear division and can display the impedance model and impedance angle at the same time.

Considering the need of real time and image in log-polar coordinate having scale and rotate un-variance property,a fast IR image registration algorithm based on coordinate dual-axis projection is proposed.

In the shape matching of the two targets similarity comparison,the focus of this approach is to use the shape context algorithm to calculate log-polar coordinate histogram Cost value,the calculation of similarity to the pre-set thresholds to determine whether his face matcheds.

Mutual information image registration based on log-polar transform

The log-polar transform (LPT) is utilized and an anti-rotation and anti-scale image matching algorithm is proposed based on the image edge feature point extraction.

Secondly,by using the scaling and rotation invariance of log-polar transform,the changes of object scale and rotation are converted into transformation in orthogonal coordinate;.

This paper presents a fast log-polar transformation(LPT) algorithm,which speeds up processing from Cartesian coordinates to log-polar coordinates as well as speeding up image matching by using a dual-axis projection algorithm.

The proposed algorithm is as follows:SUSAN algorithm used to obtain the feature corner points(interesting points) of the two images;log-polar transformation and projection correlation matching algorithm used to acquire the registration parameter of the two images.

Next,interpolated by pseudo-log-polar discrete Fourier transform(DFT),the log-polar DFT values of images can be approximated more accurately than by conventional techniques.