Elegant Solutions for Fingerprint Image Enhancement

Abstract

This invention includes image processing techniques directed to achieve feature enhancement and background-foreground enhancement in fingerprint images. The image is divided into plural segments depending on the ridge-valley directions. Each segment is separately filtered with a directional filter generally perpendicular to the corresponding ridge-valley direction. Background-foreground detection employs edge detection to identify edge pixels. These edge pixels are averaged to determine a threshold. The threshold is applied to the original image to determine background and foreground pixels. The background and foreground pixels are filtered via a watershed fill filter with separate connectivity for background and foreground pixels.

1 . A method of filtering a fingerprint image comprising the steps of:
 receiving an input fingerprint image;   separating the input fingerprint image into plural, distinct segments, each segment having ridge-valley lines within a predetermined range of angles;   directionally filtering each of said segments with a corresponding directional filter generally perpendicular to said predetermined range of angles of said segment;   assembling said directionally filtered segments into a reassembled image; and   outputting said reassembled image.   
   
   
       2 . The method of filtering a fingerprint image of  claim 1 , wherein:
 said step of directionally filtering each of said segments includes for each segment
 grouping pixels into groups parallel to the ridge-valley lines; 
 averaging pixel values of pixels within each group, and 
 replacing each pixel in each group with said averaged pixel value of said group. 
   
   
   
       3 . The method of filtering a fingerprint image of  claim 1 , wherein:
 said step of directionally filtering each of said segments includes for each segment replacing each pixel value with a weighted sum of surrounding pixel values, each weighted sum favoring neighboring pixels along said ridge-valley line of said segment and disfavoring neighboring pixels perpendicular to said ridge-valley line of said segment.   
   
   
       4 . The method of filtering a fingerprint image of  claim 1 , further comprising the steps of:
 after said reassembling step, thresholding said fingerprint image by
 detecting all ridge-valley edge pixels, 
 averaging pixel values of said detected edge pixels, 
 setting each pixel in said reassembled image to foreground if said corresponding pixel value is greater than said average edge pixel value, and 
 setting each pixel in said reassembled image to background if said corresponding pixel value is less than said average edge pixel value. 
   
   
   
       5 . The method of filtering a fingerprint image of  claim 4 , further comprising the steps of:
 after said thresholding step, watershed filling said background pixels.   
   
   
       6 . The method of filtering a fingerprint image of  claim 5 , wherein:
 said step of watershed filling background pixels includes
 for each background pixel identifying other background pixels horizontally or vertically adjacent, and 
 replacing all said identified background pixels with foreground pixels if said identified background pixels are less in number than a predetermined number. 
   
   
   
       7 . The method of filtering a fingerprint image of  claim 4 , further comprising the steps of:
 after said thresholding step, watershed filling said foreground pixels.   
   
   
       8 . The method of filtering a fingerprint image of  claim 7 , wherein:
 said step of watershed filling foreground pixels includes
 for each foreground pixel identifying other foreground pixels horizontally, vertically or diagonally adjacent, and 
 replacing all said identified foreground pixels with background pixels if said identified foreground pixels are less in number than a predetermined number.