An image generation device that enhances visual recognizability between a recognized 3D (three-dimensional) object and a substitute image synthesized in an area of a 3D object image which is a photographed image of the 3D object includes an image synthesis section that recognizes a 3D object present in the peripheral area of a vehicle and outputs 3D object attribute information indicative of attributes of this 3D object, determines a 3D object image area as an image area of the 3D object in the photographed image based on position information included in the 3D object attribute information, outputs at least one of a substitute image of the 3D object applied with a color based on the color information and a substitute image of the 3D object under a specified directional posture of the 3D object specified based on type information and directional posture information and generates a bird's-eye view image with a substitution image in which the substitution image outputted from a substitution image output section is synthesized at the position of the 3D object image area.