Automatic extraction of foreground objects from Mars images

A novel method is proposed to automatically extract foreground objects from Martian surface images. The characteristics of Mars images are distinct, e.g. uneven illumination, low contrast between foreground and background, much noise in the background, and foreground objects with irregular shapes. In the context of these characteristics, an image is divided into foreground objects and background information. Homomorphism filtering is first applied to rectify brightness. Then, wavelet transformation enhances contrast and denoises the image. Third, edge detection and active contour are combined to extract contours regardless of the shape of the image. Experimental results show that the method can extract foreground objects from Mars images automatically and accurately, and has many potential applications.