Focus enhancing method for portrait in digital image
View Patent ↗A focus enhancing method for a portrait in a digital image is applied to an electronic device capable of receiving or reading a digital image and executing a focus enhancing process with respect to a portrait in the digital image, wherein the focus enhancing process includes a foreground definition procedure for defining a head region and a body region of the portrait as a foreground of the digital image, a foreground and background segmentation procedure for cutting away an image other than the foreground from the digital image and defining the result as a background of the digital image, and a foreground and background blending procedure for blurring the background, feathering a transition region coupled to the foreground and the background, and blending the foreground, the transition region and the background to form a new digital image with a prominent portrait.
1. A focus enhancing method for a portrait in a digital image, comprising:
a foreground definition procedure, for computing a face position in the digital image by using a face detection technique, locating a detected position of a face feature according to the face position by using a feature detection, correcting the face feature according to the detected skin color by using a skin detection technique to define a head region, computing a portrait including the head region and a body region by using a dimension scale of the body region and the head region, and using the portrait as a foreground of the digital image; wherein the skin detection technique uses the detected position of the face feature as a start point to find a boundary position of a skin, and uses mean value of boundary positions of a plurality of upper and lower rows of the face feature as the boundary position of the skin for correcting the corresponding face feature;
a foreground and background segmentation procedure, for cutting an image other than the foreground from the digital image by using a graph cutting technique, and using the result as a background of the digital image; and
a foreground and background blending procedure, for blurring the background, feathering a transition region that couples the foreground and the background, such that the transition region has a smoother and more natural image, and blending the foreground, the transition region and the background to form a new digital image with a prominent portrait.
2. The method of claim 1 , wherein the foreground definition procedure further comprises a head region definition procedure, and the head region definition procedure comprises:
defining a face feature region according to a region covered by the face feature; expanding the face feature region by using a predetermined dimension scale according to a correction result of the face feature to define a head edge region; and
expanding the head edge region by using a predetermined dimension scale to define a head transition region.
3. The method of claim 2 , wherein the foreground definition procedure further comprises a body region definition procedure, and the body region definition procedure comprises:
defining a body edge region according to the head edge region based on the dimension scale of the body region and the head region;
defining a body transition region by expanding the body edge region to a predetermined dimension scale; and
using the head edge region and the body edge region as the foreground, and the head transition region and the body transition region as the transition region between the foreground and the background.
4. The method of claim 3 , wherein the foreground and background blending procedure comprises feathering the transition region coupled to the foreground and the background, and using the foreground and the out-of-focus background as a reference to simulate an out-of-focus effect of the transition region by using a Gaussian fuzzy computation, such that the image in the transition region can show a gradient effect between the foreground and the out-of-focus background.