Project

Kernel Based Approaches for Context Based Image Annotation

Exploitation of contextual information has become very important for any automatic image annotation system. In this work a method based on a kernel and keyword propagation technique is proposed for learning the image semantics for automatic annotation with a set of keywords for each image. The similarity between the images is calculated byHellinger’s kernel or RBF kernel and the images are labelled with multiple keywords using the contextual keyword propagation method and the results obtained by usig the two kernels are analysed.The annotation results obtained were validated based on confusion matrix and were found to have good  accuracy and the design is comparatively simple compared to other methods.