A NOVEL DEEP LEARNING FRAMEWORK FOR FINGERPRINT IMAGE ENHANCEMENT AND ACCURATE RECOGNITION

Abstract –This study presents a novel hybrid approach to fingerprint recognition, combining frequency domainfiltering and deep learning techniques to enhance fingerprint quality and improve recognitionaccuracy. The proposed method first applies frequency domain enhancement to reduce backgroundnoise and clarify essential fingerprint features such as ridges and valleys. Subsequently, aconvolutional neural network (CNN) is employed to further refine and correct the binary fingerprintimage, addressing any residual misrepresentations. This approach leverages the strengths of bothtraditional image processing and modern deep learning, resulting in a more accurate and reliablefingerprint recognition system, particularly in noisy or distorted conditions. Experimental resultsdemonstrate the effectiveness of this method in improving fingerprint image quality and recognitionperformance, making it a valuable solution for biometric security applications. Keywords: Fingerprint Recognition, CNN, Image Processing, Biometric Security