|
AUTHOR(S): S. Selvakani, K. Vasumathi, K. Dhashna
|
TITLE |
 PDF |
ABSTRACT Fingerprint compression is a crucial aspect of biometric systems, designed to reduce memory requirements while preserving the essential features necessary for accurate recognition. Compression can be either lossy, where some image degradation occurs, or lossless, which retains the original quality. Traditional compression methods like JPEG (using Discrete Cosine Transform) and JPEG 2000 (using Discrete Wavelet Transform) fall under lossy frequency domain techniques. However, these methods often fail to balance high compression ratios with quality retention. To address this, a spatial domain technique based on sparse representation is proposed. This method divides the image into 20x20 pixel patches and constructs a dictionary to eliminate redundancy. Patch values exceeding a predefined threshold are retained in the dictionary, while others are discarded. Performance metrics such as PSNR, MSE, and compression ratio demonstrate significant improvements over classical methods. Fingerprint images, crucial in legal and forensic investigations, often contain vast amounts of data. These images are rarely perfect and can suffer from degradation due to variations in skin texture or impression conditions. Image enhancement techniques are applied to improve minutiae detection reliability before compression. Sparse representation-based compression involves creating a dictionary of predefined fingerprint patches, dividing the image into smaller blocks, and calculating sparse coefficients for each block. These coefficients are then quantized and encoded. Experimental results on various fingerprint datasets reveal that this method is more efficient than other compression techniques, effectively preserving critical minutiae features even after compression. |
KEYWORDS Fingerprint compression, Biometric systems, Sparse representation, Lossy compression, Dictionary construction, Image enhancement, Minutiae detection, Compression ratio, DWT (Discrete Wavelet Transform), Reliability |
|
Cite this paper S. Selvakani, K. Vasumathi, K. Dhashna. (2025) Sparse Representation-Based Fingerprint Compression. International Journal of Signal Processing, 10, 4-12 |
|
