Image Specific Similar Target Detection in Aerial Images Using Gaussian Mixture Model
N Divya · 2019
Image specific similar target detection is an image analysis method, where the regions of an image similar to that of the template image or query image (mostly selected from the same image) are detected. The proposed method makes use of Gaussian Mixture Model (GMM) for modeling the template image. The quantized color features of the template image are used in building template's GMM. Based on GMM, local descriptors of the query image are extracted by ranking the Gaussian mixtures, analyzing the regional distributions, contour shapes, etc of the individual Gaussian mixtures. A novel method of conditional merging of two top ranked Gaussian mixtures is proposed to improve the similar target detection accuracy. The proposed algorithm is rotation invariant and the experimental results show that the proposed algorithm performs well for similar target detection applications.