For accurate scene-change detection, the amplitudes of sub-band signals subjected to the spectral decomposition are detected, and the detected amplitudes are smoothed and sampled during smoothing. Then, the amount of spectral change is determined based on the smoothed sub-band amplitude and the smoothed sub-band amplitude which is one sampled amplitude before that smoothed sub-band amplitude. It is then determined whether or not the determined amount of spectral change is greater than a predetermined threshold, and processing is performed based on the determination result.