Original scientific paper
https://doi.org/10.32985/ijeces.17.6.4
Image Steganography Using Complex Blocks and an Enhanced Pixel Indicator Technique With LSBMR
Sami Ghoul
; Universiti Kebangsaan Malaysia (UKM), Faculty of Information Science and Technology, Center for Cyber Security Bangi, Selangor, Malaysia. Department of Computer Engineering Faculty of Engineering, University of Zawia, Libya
*
Rossilawati Sulaiman
; Universiti Kebangsaan Malaysia (UKM), Faculty of Information Science and Technology, Center for Cyber Security Bangi, Selangor, Malaysia
Zarina Shukur
; Universiti Kebangsaan Malaysia (UKM), Faculty of Information Science and Technology, Center for Cyber Security Bangi, Selangor, Malaysia
Faizan Qamar
; Universiti Kebangsaan Malaysia (UKM), Faculty of Information Science and Technology, Center for Cyber Security Bangi, Selangor, Malaysia
* Corresponding author.
Abstract
Region-based steganography conceals data in visually complex areas of an image to minimize perceptual distortion. Traditional approaches typically rely on statistical metrics such as standard deviation or entropy to identify these regions; however, these measures often fail to capture spatial relationships among pixels and can be significantly affected by outliers, leading to unreliable complexity estimations. This paper introduces a novel block complexity measure based on local pixel similarity, which effectively captures spatial structure and enables accurate selection of embedding regions. The method segments the image into N×N blocks, quantizes pixel intensities into L levels, and calculates a similarity-based complexity value. A threshold is then applied to select the most complex blocks for data embedding. Additionally, an enhanced Pixel Indicator Technique (EPIT) is proposed to integrate repetition-based encoding with least significant bit matching revisited (LSBMR), reducing visual distortion and increasing robustness. By embedding data only in highly complex blocks, the proposed approach achieves strong imperceptibility and robustness against statistical steganalysis. Experimental evaluations on grayscale and RGB images confirm its effectiveness, achieving peak signal-to- noise ratio (PSNR) values up to 56.57 and weighted PSNR up to 86.60 for grayscale images, and exceeding 52 dB PSNR and 0.98 SSIM at 8 KB payloads for RGB images. These results demonstrate strong visual fidelity and resilience against RS and Chi-square attacks, as well as moderate resistance to convolutional neural network-based steganalysis (XuNet) and ensemble classifier-based steganalysis.
Keywords
Image steganography; Region-based steganography; block complexity; Pixel Indicator Technique;
Hrčak ID:
347896
URI
Publication date:
15.6.2026.
Visits: 42 *