Scene Change Detection Using DCT Features in Transform Domain Video Indexing

The large volumes of visual data in future multimedia databases are expected to be stored in the compressed form. The transform domain techniques are generally based on DFT (discrete Fourier transform), KLT (Karhunen-Loeve transform), DCT (discrete Cosine transform), and subbands/wavelets. We address the problem of moving object segmentation or scene change detection and also to evaluate approaches of change detection in compressed domain standards using DCT features. Our paper presens an approach to scene change detection using DCT coefficients. Furthermore, the DCT coefficients are used to detect only necessary regions of interest for subsequent image processing steps.