The Realization And Improvement Of The Basic Algorithms Of Image Restoration In Fingerprint Pattern Recognition System

The key technologies of the fingerprint pattern recognition system lie in the digital imagepre-processing technology, characteristic point extraction and recognition technology, and soon. An important condition in pre-processing stage we have to face is usual_Ly the blurredfingerprint digital images encountered in the fingerprints col_Lection. For some moreuniversal significance distortion degraded reasons in theory and practice of investigation, Theauthor researches the key steps and core algorithms of linear methods usual_Ly used in thefiel_D of digital image restoration in this article, and compare between the advantages anddisadvantages from the two aspects of subjective and objective. And then, put forward severalideas of improvements in their foundation and realize the algorithm, thus make the recoveryresults better. The fingerprint image in each of the subsequent processing stages is to bemaintained a certain degree of reality and objectivity.In this paper, it simply states the current research status, the existing problems, thepurpose and the significance of the selected topic firstly. Then it elaborates and introduces therise and development of digital fingerprint, digital image processing and relevant backgroundknowledge of computer fingerprint identification system, and introduces several necessaryconcepts and appropriately expands, thus lays the foundation for the latter algorithmsdescription.Afterwards, this paper analyzes and realizes restoration of blurred images by inversefiltering and Wiener filtering method. In view of the " abnormal state " of the traditionalinverse filtering algorithm–the denominator mathematical can not be zero, the paper deemsthe two-dimensional discrete fingerprint image to the matrix form of constitute pixels by acertain scale, and use the mathematical idea of matrix singular value decomposition, proposesa new algorithm--using two order derivative of singular value number recombinant order toestimae the blurred fingerprint image point spread function, and the third order derivative isused to remove noise; for Wiener classic filtering algorithm, using the partition idea andBayesian decision theory combined with posterior estimation method,proposes a new improved algorithm. Improved algorithms use these two methods were simple reversefiltering and classic Wiener filtering method; throngh mathematical analysis and examples ofthe effect images before and after processing in two aspects of these traditional methods andimproved algorithm,the paper compares them in the intuitive level and the abstract level.In subsequent processing, we often encounter geometric distortion for image restorationtechnology in the pre-treatment stage even more. This paper firstly analyses a variety ofcommon geometric distortion in the form of the fingerprint image, and through processing ofmathematical modeling, summarizes them as a mapping model. Then the paper analyses andresearches the traditional algorithms-the neighbor pixel interpolation and bilinearinterpolation algorithm, and uses Cubic function trilinear interpolation fitting Sinc functionapproximation method to optimizate it.In this paper, several improved algorithm are put forward, and strive to make that havevarying degrees of degradation fingerprint degraded image distortion in the pre-processingstage can be as low as possible. Thereby lay a solid foundation for later image extraction,comparison and matching. Because the blurred image distortes much more complicated thangeometric distortion,this article’s first portion is more detailed than the second portion on thelayout, namely blurred fingerprint image processing portion describes relatively meticulous.