Digital image processing

Digital image processing is mainly divided into two fields, and the processing operation in spatial domain is introduced here.

Spatial domain: processing according to the spatial representation f(x, y) of image metadata.

Transform domain: firstly, transform the spatial representation f(x, y) of image metadata, and then process the transformed data.

Transformation domain processing must have forward transformation and inverse transformation.

Spatial domain processing methods include:

Single-band operation, point operation and neighborhood operation

Binary operation, algebraic operation and contour operation.

Pixel-by-pixel calculation of a single image.

Histogram expansion (stretching)

Usually used for overall image enhancement.

In order to highlight the target or gray range of interest in the image, suppress those gray areas that are not interested.

It is generally used for image local gray enhancement segmentation stretching.

There are many methods such as exponential transformation and logarithmic transformation (enhancing the dark part and compressing the dark part).

Generally used for image local gray enhancement: Gaussian stretching on both sides and middle compression.

Square root stretching Dark places are stretched and bright places are compressed.

Spatial operation image cropping rules cropping irregular cropping image stitching

Band operation band extraction band superposition

Image cropping keeps the interesting parts and deletes other areas according to our needs.

If it is not the final operation, it needs to be slightly larger than the actual research area as much as possible.

The biggest feature of image stitching is seamless stitching. Put multiple adjacent remote sensing images together.

Precautions:

1. When overlapping, the images to be mosaicked should have similar characteristics, such as tone and texture.

2. Try to choose the splicing line at the junction of natural features with obvious differences, such as rivers, ridges and roads.

3. Color the image evenly near both sides of the suture.

Image stitching steps:

Band extraction extracts a specific band from a multi-band image file as an independent file.

Band superposition makes record group of different bands in the same geographical range into a multi-band file.

Reverse AND OR XOR operation

Binary morphology

Gray morphology

Corrosion expansion open operation and closed operation

Corrosion: it is a process of eliminating boundary points and shrinking boundaries inward, which can be used to eliminate small and meaningless targets.

Open operation: the operation of etching and then enlarging the image using the same structural elements.

Closure operation: the operation of expanding an image with the same structural elements and then etching it.

Etching: etching the input image f with structural element B.

topping

Open operation: corrosion first, then expansion.

Closed operation: expansion first, then corrosion.