Loading

Module 1: Image Filtering and Classification

Study Reminders
Support
Text Version

Set your study reminders

We will email you at these times to remind you to study.
  • Monday

    -

    7am

    +

    Tuesday

    -

    7am

    +

    Wednesday

    -

    7am

    +

    Thursday

    -

    7am

    +

    Friday

    -

    7am

    +

    Saturday

    -

    7am

    +

    Sunday

    -

    7am

    +

Spatial Filtering of Images
The spatial filtering technique is also called Local Operation, Convolution Technique, or Convolution Filtering Technique. It is used, after image enhancement, to further smoothen the image. The Two Categories of Spatial Filters that are known in the processing of remote sensing data are: Low pass filter: This emphasises regional spatial trends and de-emphasises local variability High pass filter: This emphasises local spatial variability In Remote Sensing, it is useful to compute multivariate statistical measures such as covariance or correlation among several bands to determine how the measurements co-vary. Band rationing and Principal Component Analysis are two techniques used in performing multivariate image statistics measures. Fourier Transformation Technique
Like spatial filtering techniques, the main application of the Fourier Transformation or Frequency Domain Filtering technique is to enhance images. Fourier Transformation enables a certain group of frequencies and directions to be emphasised or suppressed by algorithms known as filters.
Filtering can be implemented through Fourier Transform when it is said to operate in the frequency domain. It can also be implemented in the spatial domain of an image itself by a process called convolution.
Most image processing are implemented in the spatial domain because of the number and complexity of computations required in the frequency domain.
Unsupervised Image Classification
Image classification is the science of converting remote sensing data or images into meaningful categories representing surface conditions or classes. The Two Types of Image Classification are: Supervised Image Classification, Unsupervised Image Classification. Supervised Image Classification In Supervised Image Classification, there is human intervention, and prior knowledge of the area where the satellite is focused is required. If there is confidence in selecting training sets in supervise classification, then it possible that supervised classification can be achieved with good accuracy. In Supervised Image Classification, training sets are chosen manually for various classes. It is always a good practice that at least 2 or more training sets for each class must be selected within that image. Each training set or class results in a cloud of points which represents the variability of different pixel spectral characteristics or signatures in that class. Clustering algorithms look at the “clouds” of pixels in spectral “measurement space” from training areas to determine which “cloud” a given non-pixel falls in. The clustering algorithms that can be used in Supervised Image Classification are: Minimum Distance to Means Classification (Chain Method), Gaussian Maximum Likelihood Classification, Parallelepiped classification