A film mode detection apparatus and a method thereof can determine whether an input image is a film image by detecting existence/nonexistence of a caption in an input image field and using information of the remaining area of a field except for a caption area. The apparatus includes a caption detection unit to detect minimum values of difference values of information between successive fields from a field area including a caption area and a remaining field area excluding the caption area in an input image, respectively, and to judge whether a caption exists in a present field according to variances of the detected minimum values, a pattern generation unit to generate a pattern of a certain area of the field using the difference values of the information between the successive fields according to whether the caption exists, and a decision unit to determine whether the input image is a film image using the generated pattern. Accordingly, deterioration of picture quality due to frequency on/off operation of a film mode caused by an error of film mode judgment can be prevented.