Image Processing pt.2
The goal of image preprocessing is improving the image data that will suppress unwanted distortions and on the other hand enhance features important for further processing. Some of the methods in preprocessing are:
- background subtraction,
- mean, etc.
The user should be cautious by using enhancement techniques because they can emphasize image artifacts, or lead to a loss of information if it`s used wrong. There are several techniques to noise removal, and they are the following:
- low-pass, high-pass, band-pass spatial filtering
- mean filtering
- median filtering
When you have an image you might want to change it to fulfill special needs. For example, when you take a photo of person and it is not focused properly. You might not want the person to merge with the background. So you can sharpen the edges between different objects in the photo. On the other hand, you might want to blur a picture so others cannot recognize it anymore. To do such changes you apply a filter on your image.
There are not only different change styles (visible for the eye), but also different ways to apply them. There are two main methods to do so:
- Changing every pixel of the picture
- Changing the frequencies of the picture
Using the first method, you have to change every pixel manually. For example, if you want the image to become lighter, you have to change every pixel by increasing its value.
Using the second method, one frequency change can affect the whole image. So you can, for example, change the brightness of the image easier. On the other hand, you have to transform the image into its frequencies first. This can be done by the Discrete Fourier Transformation. This transformation is often used for image transformations.
The different changes of a picture are called image filtering. Nowadays, these "filters" are special features for digital cameras and mobile phone camera apps. For example, you can take photos which look like they were taken during motion; or like they are a drawing; or like they are blueish. All these changes were made by image filtering.
Image segmentation is the part of computer vision, that divides an image into multiple regions based on pixel colors, their intensity or texture. All similar pixels are connected in segments and are ready to be analyzed in the following up processes.
Depending on the wanted data many image segmentation algorithms and techniques are used. The result can be everything from a simple black and white or grayscale image (Thresholding) to a number of clusters defined based on pixel attributes (Clustering methods). There are also many edge detecting methods which can later be used to create 3D objects.
A wisely chosen and well executed image segmentation process is the Basis for a good and fast analysis process in computer vision and all connected applications like:
- content-based image retrieval
- machine vision
- medical imaging
- object detection
- recognition tasks
- traffic control systems
- video surveillance