Resampling Method

Introduction to Resampling Methods

SuperMap ImageX Pro provides eight resampling methods, as detailed in the following table:

Resampling Method Description
Gaussian Kernel Performs weighted mean on the entire image. Each output pixel value is derived from weighted mean calculations of itself and neighboring pixels. Particularly effective for high-contrast images with distinct pattern boundaries.
Bilinear

For any pixel p(x,y) located between four pixels fi,j, fi,j+1, fi+1,j, fi+1,j+1, the bilinear interpolation is: p=(1-dx)(1-dy)fi,j+dx(1-dy)fi,j+1+dy(1-dx)fi+1,j+dxdyfi+1,j+1, where dx=x-Int(x), dy=y-INT(y), and INT denotes the integer function.[1]

Advantages: Simple and moderately accurate, generally producing satisfactory results. Disadvantages: Exhibits low-pass filtering characteristics, potentially losing edge or linear information and causing image blurring.[1]

Cubic Assigns weighted mean values of 16 nearest input pixels to corresponding output pixels. Produces the sharpest results with edge enhancement, but requires intensive computation and longer processing time.
Nearest Directly adopts the value of the closest pixel to (x,y) in the input image. The nearest neighbor method is simple and maintains original pixel values. However, it may create discontinuities in output images and cause significant errors when adjacent pixels have large value differences.[1]
Average Uses the mean of all valid values as the resampled value.
Average Complex Data Averages magphase values in magphase space, specifically designed for resampling complex data images.
Cubic Linearity Based on Akima interpolation algorithm. This method requires four adjacent measured points beyond the two interpolation points, considering derivative effects to produce smooth and natural interpolation curves.
Lanczos Sine Resampling Utilizes convolution filters by shifting the convolution function origin to each resampling center, then multiplying and summing all input values with corresponding convolution function values.

References

[1] Wei Yuchun, Tang Guoan, Wang Min, Yang Xin, et al. Tutorial on Remote Sensing Digital Image Processing[M]. Beijing: Science Press, 2019.4:134-135.