Automatic synthetic background defocus for a single portrait image.

Capturing image with defocused background by using a large aperture is a widely used technique in digital single-lens reflex (DSLR) camera photography. It is also desired to provide this function to smart phones. In this paper, a new algorithm is proposed to synthesize such an effect for a single portrait image. The foreground portrait is detected using a face prior based salient object detection algorithm. Then with an improved gradient domain guided image filter, the details in the foreground are enhanced while the background pixels are blurred. In this way, the background objects are defocused and thus the foreground objects are emphasized. The resultant image looks similar to image captured using a camera with a large aperture. The proposed algorithm can be adopted in smart phones, especially for the front cameras of smart phones. [ABSTRACT FROM AUTHOR]