A system or the like capable of detecting lane marks more accurately by preventing false lane marks from being erroneously detected as true lane marks. A vehicle-use image processing system (100) allows a “road surface cluster” to be extracted from the “histogram” of luminance of each pixel in a “reference area” in a road surface image. Among “primary lane mark candidates,” those that overlap the “reference area” are detected as “secondary lane mark candidates.” Among the “secondary lane mark candidates,” those that have “luminance parameter” values falling within the luminance range of the “road surface cluster” are not detected as true lane marks. Thereby, lane marks are prevented from being erroneously detected (erroneous detection). This allows only lane marks to be detected more accurately.