Research On Segmentation Algorithm Of Fingerprint Images

In recent years, automatic fingerprint identification technique has been a research focus in the area of science. Many researchers have done much work on fingerprint identification. But some technical difficulties always baffle the development of fingerprint identification. Fingerprint segmentation is an important step in fingerprint recognition and is usually aimed to identify non-ridge regions and unrecoverable low quality ridge regions and exclude them as background .Remaining ridges ,which are the afterimages of the previously scanned finger ,tend to decrease accuracy of feature extraction with spurious minutiae produced enormously.Therefore remaining ridge regions should be excluded from the foreground. The valid fingerprint segmentation algorithm has important meaning to improving the speed and the performance of system.In this paper,we deeply analyze and investigate the key problems of fingerprint segmentation. The main contents include two parts:The first part: Actually there is distinct boundary between foreground and background areas of most fingerprint images,and as a consequence,combining the shape with gradient information of fingerpring images,we try to put forward a fingerprint segmentation method based on improved snake modelto remove background areas. The oval curve defined as initial outline is scattered control points by radian degree firstly. Then a energy function acting on these control points is defined and will push initial curve towards energy function descrease. The final outline didn't attain until the energy function let up.The pixels inside the outline line are set as original grey degree value, while the pixel outside the outline line as the grey degree value 255.The second part: Constantly, remaining ridges are of clear structure.Therefore, it is difficult to distinguish remaining ridges from the interested ridges by just one-step segmentation which is the typical scheme based on fingerprint features.The features used for segmentation mainly include gray-scale statistical features, local directionality, statistics of direction image,consistency of orientations, ridge projection so on. This paper proposes two steps for fingerprint segmentation aimed at removing the remaining ridge region from the fingerprint images.The non-ridge regions and unrecoverable ridge regions are removed as background by existing methods in the first step, and then the secondary segmentation is proposed to remove the remaining ridges. According to the location relationship between the remaining ridges and the real ridges , remaining ridge regions are divided into three cases: separate, conglutination and overlapped.The appropriate segmentation method for each case is proposed and used to remove the remaining ridge regions.Generally,if the remaining ridge and real ridge regions are separate to each other,the area of remaining ridge regions is smaller than real ridge regions.We can calculate each connected areas to remove separate remaining ridge regions by line label code.Under normal circumstance, the boundary outline of the fingerprint images present convex.In the case of conglutination ,the shape of boundary is abnormality. After obtaining the outline curve through the boundary track ,we use boundary point slope and curvature to find the optimal segmentation point to remove conglutination remaining ridges.While remaining ridge and true ridge regions are exactly overlapped. the former ridges directional difference round the particular point (usually singular point) is larger than the latter. we distinguish ridges directional difference of various regions to remove overlapped remaining ridge regions.