Fractal coding of bio-metric image for face authentication

Fractal objects have patterns, convergence, determinism, and reduction in dimensionality. Fractal geometry is looking for self-similarity, self-similarity resonance, and self-similarity convergence. As a result, Fractal Geometry has evolved into a scientific discipline with high predictability. Benoit Mandelbrot established the concept of the natural fractal object, characterized by merging the basics of self-similarity, scaling correlation, and statistical components. Because of these self-features, the fractal image coding method is a more dependable alternative to choose in image coding schemes. Due to this adoption, fractal image coding has already established numerous significant applications in image biometrics, picture compression, image signature, image watermarking, image extraction, and even image texture segmentation. Despite this, the fractal encoder’s popularity with the partition iterated function system rapidly drops due to its long encoding period. However, the existing strategy, which comprises two major phases and is based on the partition iterated function system, can be altered. The first stage involves encoding a statistically self-similar input image into the fixed point of an IFS, and the second involves decoding the IFS data to obtain the fractal image. Unfortunately, in existing methods, the collection of IFS data in the first step synchronizes badly, which results in lousy image quality in the decoding step. So both time complexity and image quality are not suitable for biometric face authentication. However, with these difficulties, fractal coding does not apply to personal biometric authentication unless it is resolved at a certain optimized level. This information loss in image and extended encoding times was mitigated by proposing an appropriate fractal coding technique for a biometric image. This thesis examines how fractal image coding is used in biometric cryptography for personal face authentication. This thesis implements the algorithms for the Methods of CPM, BPBM fractal coding, and its application. This thesis proposes a novel way of integrating Fractal coding into a digital smart card with embedded cryptographic encryption and decryption and guard against all assaults, according to cryptoanalysis research. The thesis gains information on the time required to encode the biometric image by implementing techniques of CPM and BPBM and measures their impact on the encoding duration. The thesis also compares the results of enough images of various sizes generated by the proposed algorithms with the results of other fractal coding methods to confirm the algorithms’ clarity, reliability and validity. Finally, this thesis will apply BPBM methods in biometric face authentication. The thesis finds that there are core gaps and plans accordingly based on the literature review. The CPM and BPBM have been proposed, with two main streams: encoding and decoding of both. In the encoding, the fractal function converges to its self-similarity as IFS. The inverse function calls IFS back to create a corresponding fractal object in the decoding stream. The first method (CPM) blueprints blocking dimension, pooling factor, method, and block matching. The second method designs (BPBM) Pixel Binarization. Both methods, CPM and BPBM, are clarified, justified and validated with experiments and Benchmarks. Finally, BPBM is considered implementing biometric face authentication. In CPM, the odd size pixel dimension of blocks, odd size pooling and max pooling scheme for smoothing, and the entire domain blocks search by only a single central pixel are better redeemable aspects throughout the encoding phase by adjusting to the previous idea. The key implication of these principles is that for both blocks, the symmetrical central pixel is used to search the relevant domain block rather than the entire neighborhood of the block. This study determined that this symmetrical central pixel will conform to the best-fit domain block by matching the same range block. In BPBM, the thesis contributes to search space reduction by converting eight-bit pixels to two-bit pixels using the central pixel value of blocks and the related eight-neighbors. As a result, the thesis contributes that excess block space is shortened before the search begins. Because of the limited bits of domain blocks, we achieve faster coding by reducing run-time compared to any current exhaustive block hunting approach. This additional investigation encourages the improvement of encoding speed and afterward reveals enhanced results employed in personal authentication. We assess the study results in three essential areas: encoding time complexity in second, fractal object (image) quality in PSNR, SSIM, FSIM (seventeen quality features have been used), dimensionality reduction (Compression Ratio). In all aspects, CPM and BPBM confirm the superiority.

Text MD AHADULLAH - IR.pdf Download (975kB)

Actions (login required)

View Item