Comparative high contrast area extraction in image based on spatial-contrast feature

Referring to the fact that people recognize object often by the contrast difference between the object and the background, this paper proposes two new features named the absolute-contrast and the spatial-contrast, while presents a new method of extracting comparative high contrast area in image based on the spatial-contrast. First, the absolute-contrast is defined as the longest distance between peaks of the image´s histogram, which is used to measure the whole contrast of the image. Second, aiming to simulate the perceive mechanism of human being, the spatial-contrast feature is proposed to describe the contrast difference between the area including the object and the entire image. Third, the algorithm of comparative high contrast area extraction based on spatial-contrast feature is developed. Finally, the method is verified by the text caption area extraction in video. The result shows that the new method is feasible and efficient. Because the spatial-contrast focuses on the feature related with regional contrast, it is qualified as a measure of salience in object recognition, which is free of the object shape and geometric transformation. So, such new features and method can provide a good basis for recognition following.