Face image processing method

• US 8,295,557 B2
• Filed: 01/12/2009
• Issued: 10/23/2012
• Est. Priority Date: 01/12/2009
• Status: Active Grant

• Current Assignee: Arcsoft Corporation Limited
• Original Assignee: ArcSoft, Inc.
• Inventors: Wang, Jin, He, Cheng-Jian, Li, Jian-Feng, Wang, Guo-Gang, Liu, Xiao-Mu
• Primary Examiner(s): Le, Vu
• Assistant Examiner(s): Allison, Andrae S
• Application Number:
  US12/318,891
• Publication Number:
  US 20100177981A1
• Time in Patent Office: 1,380 Days
• Field of Search: 382/118, 382/170, 382/190, 382/171, 382/162, 382/167, 382/165, 382/231.4, 382/222.1, 382/275, 348/169, 348/129, 348/231.4, 348/222.1, 358/501, 358/518, 358/523, 340/469, 340/425.5
• US Class Current: 382/118
• CPC Class Codes: G06T 2207/10024 Color image
• CPC Class Codes: G06T 2207/20016 Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
• CPC Class Codes: G06T 2207/30201 Face
• CPC Class Codes: G06T 5/77 Retouching; Inpainting; Scratch removal
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06V 40/162 using pixel segmentation or colour matching

Abstract

A face image processing method is applied to an electronic device, such that the electronic device can perform a face detection to a digital image to obtain a face image in the digital image automatically, and perform a skin color detection to the face image to exclude non-skin features such as eyes, eyeglasses, eyebrows, a moustache, a mouth and nostrils on the face image, and form a skin mask in an area range of the face image belonging to skin color, and finally perform a filtering process to the area range of the face image corresponding to the skin mask to filter high-frequency, mid-frequency and low-frequency noises of an abnormal skin color in the area range of the face image, so as to quickly remove blemishes and dark spots existed in the area range of the face image.

Claims

1. A face image processing method, which is applied to an electronic device, comprising the steps of:

automatically performing a face detection to a digital image to obtain a face image in the digital image;

marking a frame on the face image after the face image is obtained;

performing a skin color detection to the face image corresponding to the frame to exclude non-skin features on the face image, and forming a skin mask on the face image within an area range belonging to a skin color, wherein a multi-seeds algorithm is adopted when the skin color detection is performed to the face image, and the multi-seeds algorithm comprises the steps of;

forming a predetermined quantity of uniformly arranged seeds within the frame;

determining whether or not a pixel color at a corresponding position within the frame is the skin color according to the position of each seed, in order to perform a skin color filtering;

computing a center color for the pixel colors belonging to the skin color after the skin color filtering takes place; and

sorting the pixel colors and the center colors corresponding to all the seeds, and deleting the seeds having colors different from the center color over a predetermined chromatic aberration, and the remaining seeds become the skin mask; and

performing a filtering process to the area range corresponding to the skin mask on the face image to filter high-frequency, mid-frequency and low-frequency noises in the area range of the skin mask.

2. The face image processing method of claim 1, wherein the multi-seeds algorithm forms the seeds within an inscribed circle or an ellipse of the frame.

3. The face image processing method of claim 2, wherein the method adopts a Gaussian and Laplacian pyramid operation to separate high-frequency, mid-frequency and low-frequency noises from the pixels in the face image corresponding to the skin mask.

4. The face image processing method of claim 3, wherein the method adopts a non-local (NL) filtering process to filter the high-frequency, mid-frequency and low-frequency noises representing abnormal skin colors.