5.17.1. Image pre-processing

Image pre-processing involves the initial processing of raw image data to apply corrections for geometric distortions, calibrate the data radiometrically, and remove any noise present in the data, if any.

Image pre-processing is the first step in working with digital images, particularly in remote sensing. It includes essential corrections necessary to ensure the data is usable for analysis. The main goals are to fix any geometric distortions, adjust the data to make it radiometrically accurate, and eliminate noise that might interfere with the interpretation of the image. These steps are crucial because raw images may contain inconsistencies that are influenced by the equipment used or environmental conditions.

Think of preparing a dish before serving it. When you cook, the ingredients might not blend perfectly due to uneven heating or wrong proportions. Just like a chef tastes and adjusts the seasoning or even alters the cooking method to ensure the final dish is perfect, image pre-processing adjusts the digital images so that analysts can 'taste test' or interpret the data accurately.

Geometric corrections have two basic steps: (i) Georeferencing: Raw remote sensing data is without any geographic coordinates and has distortions, mainly caused by the sensor geometry. Georeferencing is the conversion of image coordinates to ground coordinates by removing the distortions caused by the sensor geometry. (ii) Resampling: After georeferencing, we may find that the pixels have been oriented differently than the way they were present in the original image coordinate system.

Geometric corrections are necessary to align images with actual geographic coordinates. Georeferencing is the process of adjusting the image so it matches a global coordinate system, allowing different images to be compared and analyzed accurately. This involves identifying points on the image that correspond to known locations on the Earth, called Ground Control Points (GCPs). Resampling follows georeferencing, which adjusts pixel values to align with their new positions in the corrected image, ensuring that the data remains accurate. Without these steps, any measurements or comparisons made would be flawed.

Imagine trying to fit a jigsaw puzzle piece that has been rotated or flipped. You would need to twist and turn the piece until it fits perfectly with the rest. Georeferencing is akin to un-scrambling the puzzle piece's position so it fits within the whole picture, and resampling is about ensuring that every piece has the right portion of the correct colors to match the image.

Atmospheric correction is done to modify the DN values to remove noise, i.e., contributions to the DN due to intervening atmosphere.

Atmospheric correction addresses issues where the atmosphere can distort the data captured by sensors. Factors such as haze or clouds can affect light reflectance captured in the digital image, leading to inaccurate readings. The goal of this correction is to adjust Digital Number (DN) values to account for these factors, improving the accuracy of the data. One method used for this correction is the dark object subtraction method, where the lowest DN value is identified and subtracted from all other values, thereby normalizing the data for any atmospheric noise.

Consider taking a photograph on a foggy day. The fog creates a blur, distorting the colors and shapes of the subjects in the image. To make the photo clearer, you could apply post-processing techniques that enhance the image, making colors pop and shapes more discernable. Similarly, atmospheric correction clears away the 'fog' from satellite images, revealing the true colors and details of the Earth's features.

--