Types of Resampling Methods
Resampling is the process of estimating new pixel values after geometric transformation of an image
during geometric correction. Since the corrected image grid does not perfectly match the original pixel
positions, new pixel values must be assigned using resampling techniques.
1. Nearest Neighbour (NN)
Nearest Neighbour Technique is the simplest resampling method used in remote sensing and GIS during
geometric correction and image transformation. It is used to assign values to new pixels when an image
is rotated, enlarged, shifted, or corrected.
In this method, the value of the nearest original pixel is directly assigned to the new pixel location. No
averaging or interpolation is performed. Therefore, the original pixel values remain unchanged. Nearest
Neighbour resampling is preferred when preservation of original pixel values is more important than
image smoothness. Although the output image may appear less smooth, it is highly useful for
classification studies and thematic data analysis.
Working Procedure
• Locate the new pixel position.
• Identify the nearest pixel from the original image.
• Copy its digital number (DN) value to the new pixel.
Features
• Uses only one nearest pixel
• Preserves original spectral information
• Produces blocky or jagged appearance
Advantages
• Very simple and fast method
• Maintains original brightness values
• Suitable for classified and thematic images
Disadvantages
• Image quality may appear rough
• Creates staircase or blocky effect
• Less smooth compared to other techniques
Applications
• Land Use and Land Cover (LULC) classification
• Thematic mapping
• Remote sensing image rectification
• Categorical data analysis
2. Bilinear Interpolation
This method provides smoother results than nearest neighbour. Bilinear Interpolation is a commonly
used resampling technique in remote sensing, GIS, and digital image processing. It is applied during
geometric correction, image registration, rotation, scaling, and transformation of satellite images.
When an image undergoes correction or transformation, new pixel locations are created. Since these
new locations may not exactly match the original pixel positions, bilinear interpolation estimates the
new pixel value using information from nearby pixels.
In this method, the four nearest neighbouring pixels surrounding the new pixel are considered. A
weighted average of these four pixels is calculated to determine the new pixel value. Pixels located
closer to the new position influence the value more than distant pixels
Principle
• Uses the four nearest surrounding pixels.
• Computes the new pixel value using weighted averaging.
Characteristics
• Produces smoother images.
• Slightly modifies original pixel values.
Advantages
• Better image appearance
• Reduces blocky effect
Disadvantages
• Alters original spectral information slightly
• Not ideal for precise classification wo
Applications
• General image processing
• Image enhancement
• Continuous data analysis
3. Cubic Convolution
Cubic Convolution is an advanced resampling technique used in remote sensing, GIS, and digital image
processing during geometric correction, image transformation, and image enhancement. It is used to
estimate new pixel values when an image is enlarged, rotated, shifted, or corrected.
In this method, the value of a new pixel is calculated using the weighted contribution of the 16 nearest
surrounding pixels arranged in a 4 × 4 neighbourhood. A cubic mathematical function is applied to
obtain a smoother and more accurate result.
Compared to nearest neighbour and bilinear interpolation methods, cubic convolution provides better
image quality and sharper appearance.
Principle of Cubic Convolution
• The method considers 16 surrounding pixels.
• Each pixel contributes to the final value based on its distance from the new pixel location.
• A cubic interpolation function is used for calculation.
This produces smoother transitions and clearer edges in the image.
Working Steps
1. Identify the new pixel location.
2. Select the 16 nearest neighbouring pixels.
3. Apply cubic mathematical interpolation.
4. Compute the weighted average.
5. Assign the calculated value to the new pixel.
Example
Suppose a new pixel is located within a group of 16 neighbouring pixels. Instead of selecting one pixel
or averaging four pixels, cubic convolution uses all 16 surrounding pixels to estimate the new value
more accurately.
The resulting image becomes smoother and visually clearer.
Characteristics
• Uses 16 neighbouring pixels
• Produces high-quality smooth images
• Computationally more complex
• Modifies original pixel values
Advantages
• Produces very smooth and sharp images
• Better edge representation
• Reduces distortion and pixelation
• Suitable for image enlargement and visualization
• Higher visual quality than bilinear interpolation
Disadvantages
• Slower processing time
• Requires more computation
• Original spectral values are altered
• Not suitable for classified or thematic data
Applications
Cubic convolution is used in:
• Satellite image enhancement
• High-resolution image processing
• Aerial photograph correction
• Image enlargement and zooming
• Continuous surface analysis