A novel image coding framework based on content similarity analysis

Abstract

Because of not considering the correlations among images, current image coding technologies result in an unnecessary waste of storage space. This paper proposed a novel image coding framework based on content analysis, which can help to elevate the storing efficiency by utilizing the removal of spatial correlations within the image set composed by similar images. In our method, SVM algorithm is used to classify similar images after extracting image features of BOF (Bag of Feature). All images of each category are considered as a GOP (Group of Pictures) within MPEG-2 standard, and then each of them will be further compressed with predictive coding. Experimental results demonstrate that the proposed method can improve coding efficiency by reducing redundancy of correlated images, and further greatly saving the storage space.