Joseph Fofanah

Followers

Following

Public Views

Uploads

Papers by Joseph Fofanah

Dual Watermarking for Protection of Medical Images based on Watermarking of Frequency Domain and Genetic Programming

Proceedings of the 2020 the 4th International Conference on Innovation in Artificial Intelligence

In this paper, we presents two watermarking schemes namely: Scheme I present an hybrid of discret... more In this paper, we presents two watermarking schemes namely: Scheme I present an hybrid of discrete cosine transform (DCT) and discrete wavelet transform (DWT) watermarking technique while Scheme II presents a genetic programming (GP) based technique. Both of these schemes were experimented with 512x512 and 256x256 medical images and other standard images such as Lena, Barbara, Baboon, Flower, Brain etc. against state-of-the-art watermarking techniques. According to our results, shows better performance of the two main criteria use to check for any watermarking techniques (Imperceptivity and Robustness). The experimental results reveals that for imperceptivity, peak-signal-to-noise-ratio (PSNR) and structural similarity index (SSIM) shows better performance when compared against Arsalan's and Hurrah's approaches. Similarly, for robustness the parameters used are normalized correlation (NCC) and bit error rate (BER). According to results obtained, both schemes' parameters show high performance except for Scheme I BER with 0.0038. Finally, in terms of fragility both schemes provided better resistance to any signal processing attacks.

Medical images lossless recovery based on POB number system and image compression

Multimedia Tools and Applications, 2022

To protect the information integrity of the medical images, this paper proposes a secure lossless... more To protect the information integrity of the medical images, this paper proposes a secure lossless recovery scheme for medical images based on image block compression and permutation ordered binary (POB) number system, which includes two parts: share generation, and lossless recovery. In the shares generation stage, the region of interest (ROI) of the original medical image and the image block compression coding algorithm JPEG-LS, are firstly adopted to generate the compressed data, which can use limited storage space to place the data repeatedly. Then, after separating the bit-plane, the compressed data are executed by data reorganization and data encryption respectively on the high-plane and the low-plane. Finally, two authentication bits are extracted by 8-bit pixel of two planes to be inserted into the pixel itself, and then the 10-bit pixel is converted into an 8-bit POB value to produce two shares of ShareH and ShareL, respectively. In the lossless recovery stage, data of placing repeatedly can restore the original image. At the same time, three attacks are carried in the two shares, which contain content cropping, content exchange, and text addition. Some comparisons with other schemes present that the proposed scheme implements a better performance under some criteria. Theoretical analysis and experimental results demonstrate that the original image can be recovered losslessly even if two shares are tampered at the ration more than 50%.

format_quoteLossless recovery of the original image is achievable even when two shares are tampered beyond 50%.format_quote

Download

Adaptive Neuro-Fuzzy Inference System with Non-Linear Regression Model for Online Learning Framework

International Journal of Scientific & Engineering Research, 2020

In this paper, we present a combined hybrid of an adaptive neuro-fuzzy inference system with an a... more In this paper, we present a combined hybrid of an adaptive neuro-fuzzy inference system with an adaptive online learning model (ANFIS-AOLM) using a non-linear regression machine learning technique to construct the Neural Network. We proposed ANFIS-AOLM for learning by modeling and controlling imprecise defined, uncertainty system with a significant role of the neuro-fuzzy method. The various simulation exercise was conducted with define input and output values, define rules and the dataset was created and trained the Neural Network. Exploration of the dataset was tested for ANFIS-I using &quot;Grid Partitioning&quot;; ANFIS-II using &quot;Subtractive Clustering&quot;; and ANFIS-III using &quot;Fuzzy Clustering Means&quot;. The study was conducted in oder to explore the effectiveness of the combined hybrid learning opportunities using ANFIS and Non-linear Regression models. The study was experimented using MATLAB programming language for the adaptation of the online learning framework with learner&#39;s capability to learn from the existing information and adapt accordingly in the learning environment. The results obtained from the analysis shows that with a membership function of 20 and 25 an average of 97.7% accuracy was achieved. However, as the number of membership functions increases the better the performance of the ANFIS and the higher the computational response time. The non-linear regression model also indicated that the relationship of the ANFIS system and the proposed technique is adequate for learner&#39;s adaptation and the decision made through learning mechanism.

Joseph Fofanah

Uploads

Papers by Joseph Fofanah

Log In