A Decision-Based Median Filter with Adaptive Neighborhood Logic for Efficient Noise Reduction in VLSI-Oriented Image Processing Applications

In the realm of image processing, filters are essential for mitigating various types of noise, including random, salt and pepper, and Gaussian noise. The effectiveness of these filters is crucial, particularly in real-time applications and Very Large-Scale Integration (VLSI) hardware implementations. Traditional hardware-based filters often face challenges related to lookup table (LUT) requirements, route delays, and power consumption. This research focuses on developing a Decision-Based Median Filter (DBMF) designed to address these issues through a novel approach. The proposed DBMF integrates data comparator decision-based multiplexer logic with adaptive neighborhood logic. Initially, a decision-based multiplexer, managed by selection logic, differentiates between high and low values within a pair of numbers to ensure accuracy. This process is iterated across nine-pixel combinations to compute the median value. The proposed DBMF demonstrates superior performance in noise reduction and key hardware metrics, including latency, power consumption, and LUT usage, compared to existing advanced methods. Evaluation results show that the DBMF achieves a Peak Signal-to-Noise Ratio (PSNR) of 45.84, a Mean Squared Error (MSE) of 0.00477, and a Structural Similarity Index (SSIM) of 0.9816, indicating substantial improvements in filter performance and efficiency. These findings highlight the DBMF’s effectiveness in reducing noise and enhancing overall hardware performance.