Traditional Culture Encyclopedia - Photography major - Three resampling methods

Three resampling methods

Three resampling methods: nearest neighbor method, bilinear interpolation method and cubic convolution method.

Introduction to resampling:

Resampling refers to the process of interpolating the information of one type of pixel according to the information of another type of pixel. In remote sensing, resampling is the process of extracting low-resolution images from high-resolution remote sensing images. The commonly used resampling methods include nearest neighbor interpolation, bilinear interpolation and cubic convolution interpolation.

Application field:

In digital photogrammetry and remote sensing, the actual resampling occurs in the process of image rotation, polar line queuing, digital correction and multi-image synthesis. After geometric transformations such as registration, correction and projection, the center position of the original pixel often changes, and the row number and column number of its position in the input grid are not necessarily integers.

Therefore, it is necessary to resample the input raster according to the position of each pixel on the output raster in the input raster, that is, to recalculate the value of each raster and establish a new raster matrix. This requires image resampling after registration. When calculating between raster data and image data with different resolutions, resampling is also needed.

Usually, the grid size is unified to a specified resolution, that is, the position of each pixel and the number of pixels in the same area will also change. SuperMapDeskpro .NET provides three commonly used resampling methods: nearest neighbor method, bilinear interpolation method and cubic convolution method.

Nearest neighbor method:

The nearest neighbor method takes the pixel value closest to the pixel position in the image as the new value of the pixel. This method has the advantages of simplicity, high efficiency, fast calculation speed and no change in the grating value of the original image. The disadvantage is that the maximum displacement will be half a pixel, and the calculation is not accurate enough.

It is suitable for representing discrete data of land parcel classification or thematic map, such as forest cover, land use classification, vegetation type statistics, etc. The following figure shows that the output raster is resampled by the nearest neighbor method after the raster data is geometrically transformed (such as translation and rotation), and the black wireframe represents the input raster.