Raster Resampling is the process of assigning pixel values or derived values from the input image to each pixel in the output image. The image mentioned here is Raster Data, including GRID and IMAGE. Raster Resampling is required when the position of the input image and the output image (through operations such as geometric transformation or Projection Settings) or the pixel size (i.e., Raster Image resolution) changes.
In addition, Raster Resampling is a common Data Processing method used by Raster Data to deal with the problem of raster resolution matching in Spatial Analysis. Different resolutions are usually converted to the same resolution by Raster Resampling. For the Raster Data with a given spatial resolution, the Raster Data can be resampled into larger pixels through the resampling operation, that is, the spatial resolution is reduced. This process will lose some of the details of the original high spatial resolution information; it can also be resampled into smaller pixels, but it will not add more information. The multispectral Remote Sensing Imagery with low spatial resolution is resampled to the same resolution as the Panchromatic Image with high spatial resolution, and then the two images are fused to obtain an image with both high spectral resolution and high spatial resolution information, which can be used for thematic extraction and application. It is a common way of remote sensing data fusion.
Raster Resampling mainly includes three methods: Nearest, Bilinear and cubic convolution interpolation. Nearest is to fill the nearest pixel value in the Original Image into the new image; Bilinear and cubic convolution interpolation are to fill the pixel value near the Original Image into the new image through the distance Weighted Mean. These three Resampling Methods are described in detail below.
Nearest
Nearest takes the nearest pixel value in the input Raster Dataset as the input value and assigns it to the corresponding pixel of the output Raster Dataset.
The advantage of this method is that it does not change the original grid value and the processing speed is fast, but this method will have a displacement of half a pixel size. It is suitable for discrete data representing classification or a certain theme, such as Land Use, vegetation type, etc.
 |
As shown in the figure above, the output Raster Dataset is resampled using Nearest after geometric transformations such as translation and rotation. The black line frame represents the input Raster Dataset, the light green fill represents the output Raster Dataset, and the red square dot represents the center position of a pixel in the output Raster Dataset, whose pixel value will be Recal culate. Find the center point of the pixel nearest to the pixel where the red square point is located, that is, the purple dot shown in the figure, and fill the pixel value represented by the purple dot into the red square point to complete the resampling of a grid pixel.
Bilinear
The double Linear interpolation method is a method based on cubic Linear interpolation, which calculates new pixel values by performing Weighted Mean on the pixel values of the four nearest pixels (four neighborhoods) in the input Raster Dataset. And assign it to the corresponding pixel of the output Raster Dataset. The weight is determined by the distance between the center of each pixel in the 4-neighborhood and the interpolation point.
The Resample Result for this method will be smoother than Nearest, but will change the original grid values. It is suitable for representing the continuous data of a certain phenomenon distribution and terrain surface, such as DEM, temperature or rainfall distribution, slope, etc. These data are originally the continuous surface obtained by interpolation of sampling points.
 |
As shown in the figure above, the output Raster Dataset is resampled using Bilinear after geometric transformations such as translation and rotation. The black line frame represents the input Raster Dataset, the light green fill represents the output Raster Dataset, and the red square dot represents the center position of a pixel in the output Raster Dataset, whose pixel value will be Recal culate. Take four adjacent points around the red square point, and the center points of these four adjacent points are the positions indicated by the purple dots in the figure. By calculating the distance Weighted Mean, fill the Calculate Result into the red square point to complete the resampling of a grid pixel.
Cubic
Similar to Bilinear, Cubic is a method based on quintic polynomial interpolation, which calculates new pixel values by Weighted Mean of the pixel values of the 16 nearest neighboring pixels (16 neighborhoods) in the input Raster Dataset, and assigns them to the corresponding pixels in the output Raster Dataset. The weight is determined by the distance between the center point of each pixel in the 16 neighborhood and the interpolation point.
Cubic can obtain the best interpolation function by increasing the adjacent points, which can further improve the interpolation accuracy. The algorithm is complex, the amount of calculation is large, and the processing time is long. Because this method uses 16 neighborhoods for weighted calculation, the processing results will be clearer, and the boundary of Raster Data will have a sharp effect. This method also changes the original grid values and may exceed the range of values of the input grid. It is suitable for the resampling of aerial photographs and Remote Sensing Imagery.
 |
As shown in the figure above, Cubic is used to resample the output Raster Dataset after geometric transformations such as translation and rotation. The black line frame represents the input Raster Dataset, the light green fill represents the output Raster Dataset, and the red square dot represents the center position of a pixel in the output Raster Dataset, whose pixel value will be Recal culate. Take 16 adjacent points around the red square point, and the center point of these 16 adjacent points is the position indicated by the purple dot in the figure. The distance Weighted Mean is calculated, and the Calculate Result is filled into the red square points to complete the resampling of a grid pixel.
Related topics
