Fingerprint Compression based on Sparse Representation using Pixel Level Patch Decomposition

Abstract: Reducing memory size of image is known as Compression. In the process of compression there will be degradation in the image, such type of compression is known as Lossy compression. The counter part of losssy compression is Lossless compression, where image is not degraded and quality is retained. The classical methods to compress an image were JPEG using DCT and JPEG 2000 using DWT. These methods come under frequency domain lossy compression standards. To improve the compression ratio and retain the quality of compressed image compared with these classical methods a new method in spatial domain is proposed which is known as sparse representation using pixel level patch decomposition. The decomposition is done on the image based on the pixel values at the level of 20X20 patches and a dictionary is constructed to remove the redundancy in the image, a threshold value is fixed the patch values which are greater than the threshold values retained in the dictionary and remaining are discarded. The metrics like PSNR, MSE and compression ratio are calculated and significant improvement is observed in the proposed method.

Keywords: JPEG, JPEG2000, Sparse Representation, Fingerprint Dictionary.