An apparatus for detecting a mixed area in frames. An object extracting unit extracts a foreground object from an input image and generates an area-specified object formed of the foreground object and a value indicating that the foreground object belongs to a background area. A motion compensator compensates for the motion of the area-specified object based on a motion vector and positional information thereof. A subtracting unit subtracts the pixel value of a pixel belonging to a foreground object of a current frame from the corresponding pixel of the foreground object of a preceding frame, so as to obtain a frame difference between the pixels belonging to the foreground area. A threshold-value processor detects a mixed area based on the difference.