Content Based Image Retrieval for Space Weather Application

Abstract

The research presented in this paper was aimed to develop an CBIR system to aid in automatically detect and retrieve solar flares images based on detected edges and color features. In general, the work consists of two phases: (1) enrollment phase, which consist of feature extraction based on Conventional Color Histograms (CCH) and edge detection for sunspots and active regions on the sun using Sobel filter, (2) retrieving phase, which use the similarity measurement to retrieve most similar images from the database for a given query sample. The conducted tests were carried on 250 images and give good precision and recall rates.