Fingerprint Image Processing
 

     Personal identification by means of biometrics has drawn more attention due to the needs in reliable and easy to use a system with affordable cost. Among the many possible biometric schemes, the fingerprint is one of the most reliable mean for the identification of individual since each person holds distinct pattern.

Identification by mean of a fingerprint, there are many more steps and research challenges; for instances: Image acquisition and compression, Image enhancement, Fingerprint classification, and fingerprint matching. Each can be roughly details as follows:

 

Image acquisition and compression:

Fingerprint image are officially accepted at the solution of 500 dpi (dot per inch). There are many types of sensor used in acquiring the image, namely; optical sensor, resistive sensor (thermal scan), and capacitive sensor. Each type of sensor together with the corresponding technique offers slight difference image characteristics. The achieved raw image should be compressed to increase the number of image per megabyte when the data are to be stored. However, the quality of the image must be maintained to that as of as close as the original ones. As the database becoming larger and larger, the compression technique is therefore of interest.

 

Image Enhancement:

The scanned images are frequently not clear. The unclear or low quality images lower the recognition rate. Too dry or too wet image are sometimes unavoidable. Unintended printed images (criminal cases) are usually obtained not only in portion but also poor quality. In these circumstances, the improvement (or enhancement) of the image is very important and of interest to researchers. Particular type of filtered as well as area of interest are the main focuses of the role.

 

Fingerprint Classification:

Although fingerprint verification and matching are the main interest, fingerprint classification is of essential as their pre-processing step. The first scientific study of the fingerprint was made by Galton who divided fingerprint into three major classes. Henry, later refined Galton’s classification by increasing the number of classes to five classes, namely; arch, tent arch, left loop, right loop and whorl. Henry’s classification is well-known and widely accepted. At a global level (or high level) the fingerprint pattern exhibits the area that ridge lines assume distinctive shapes. Such an area or region with unique pattern of curvature, bifurcation, termination and etc is firstly classified into three main topologies; loop, delta and whorl. The core point is defined as the top most point on the most inner ridge and a delta point is defined as the point where three flows meet. At a fine level or local level, the characteristic of ridges and valley, known as minutiae, are utilized. There are several methods for automatic fingerprint classification, namely, rule-based, syntactic, structural, statistical, neural network, and combined techniques. Each technique holds their accuracy and computational complexity. In defining the core point (and/or delta point), Poincare technique has been widely used. To detect the turning point or corner, some of them have used geometry region as well as maximum curvature technique. Fingerprint classification can be also viewed as a coarse level matching of fingerprint since it can limit the amount of searching domain. In many cases, even an expert still needs great effort to match the fingerprint. These are mostly according to only a portion of the fingerprint is available. In such case classification is also not very helpful.

 

Fingerprint Matching and Personal Identification:

Based on the obtained fingerprint image and the ones stored in the database, matching is the final step in personal identification. However as the image has to be compared with hundreds or thousands of the stored pattern processing speed as well as the correct obtained result are essentially desired. Several fingerprint featured are used at this step. The comparison can be performed in the time domain (minutiae, and other geometry structures) and/or frequency domain (as spectrum).

 

Research Methodology:

Investigated techniques and/or algorithm are to be implemented under the software simulation environment. The test images are the well-accepted free database as FVC-2002, 2004. If necessary the raw acquired images must be prepared with the commercial available scanner.

As conventional methodology; Problem identification, Literature review, hypothesis on new technique (or a modification of the existing ones), development and verification of the new proposed matter, and summary of the achievement are to be conveyed.