A REVEW OF IMAGE CLASSIFICATION TECHNIQUES BASED ON PERFORMANCE METRICS

The objective is to classify satellite images and highlight the differences either based on object, type of land cover or land usage. Image classification is the essential phase of digital image analysis. The main purpose of image classification is to automatically classify each and every
pixel in an image into multiple classes. Typically for remote sensing applications, multispectral data is utilized to perform the classification, and furthermore the spectral pattern present in the image is the basis for categorisation. This paper presents the fundamental ideas of image classification, performance metrics and comparison of various image classification techniques used by researcher in various application.The performance metrics such asAccuracy, Precision, Recall or Sensitivity, specificity and Kappa Statistic are compared and analyzed. Image
classification techniqueswere used effectively in many fields such as agriculture, remote sensing, medical imaging and so on.