A method for classification of scenery documentary using MPEG-7 edge histogram descriptor

A double-threshold method is proposed to classify the scenery documentary using MPEG-7 edge histogram descriptor (EHD). The segments of a scenery documentary are classified into two categories, building and natural scenery clips, by global EHD of shot keyframes with the double-threshold method. Thus the basic semantic of shot can be extracted that describes the building or scenery. Furthermore the precision of EHD computing has been improved by enhancement and smoothing of the pictures and the results of classification are modified under the time constraint on shot keyframes. The experiments indicate that the accuracy of classification exceeds 90%.