BFORE: Butterfly-Firefly Optimized Retinex Enhancement for Low-Light Image Quality Improvement
Low-light image enhancement has emerged as a significant challenge in the field of computer vision and multimedia applications. Images captured in poorly lit environments often suffer from issues such as poor visibility, low contrast, and color distortion, hindering their practical use in various domains. Addressing these challenges, researchers have introduced an innovative approach known as BFORE (Butterfly-Firefly Optimized Retinex Enhancement), which aims to enhance image quality through a novel framework that optimizes Retinex-based methods.
The BFORE framework operates under the premise that existing Retinex-based methods often rely on manually tuned parameters, which can be ineffective across diverse lighting conditions. To tackle this limitation, BFORE leverages a hybrid metaheuristic optimization strategy to automatically tune the parameters of a multi-stage Retinex-based pipeline. This not only streamlines the enhancement process but also improves overall image quality.
Key Components of BFORE
- HSV Color Space Conversion: BFORE begins by converting the input image into the HSV (Hue, Saturation, Value) color space. This step allows for more controlled manipulation of the image’s luminance channel, which is crucial for effective enhancement.
- Adaptive Gamma Correction with Weighted Distribution (AGCWD): The luminance channel undergoes AGCWD, which adjusts the brightness levels more effectively, accommodating varying lighting conditions.
- Adaptive Denoising: Following gamma correction, BFORE applies adaptive denoising techniques to minimize noise while preserving essential image details.
- Optimization Algorithms: The framework utilizes a Butterfly Optimization Algorithm (BOA) to optimize the parameters of the Multi-Scale Retinex with Color Restoration (MSRCR). Concurrently, it employs a Firefly Algorithm (FA) to fine-tune the AGCWD and denoising parameters.
- Hybrid BOA-FA Switching Strategy: BFORE incorporates a dynamic switching strategy between BOA and FA, effectively balancing global exploration and local exploitation, thus enhancing optimization efficiency.
Experimental Evaluation
The effectiveness of BFORE has been validated through experimental evaluations on the LOL benchmark dataset, which comprises 15 paired test images. The results indicate that BFORE achieves a peak signal-to-noise ratio (PSNR) of 17.22 dB, the highest among traditional enhancement methods. This represents a 20.3% improvement over Histogram Equalization and a 17.5% improvement over the MSRCR method. Additionally, BFORE is noted for producing a mean brightness value of 129.97, which is closest to the ideal mid-tone value, indicating a natural balance in brightness across enhanced images.
Comparison with Deep Learning Methods
Notably, BFORE outperforms RetinexNet, a prominent deep learning baseline, in both PSNR and structural similarity index (SSIM) metrics without necessitating any training data. Specifically, BFORE records a PSNR of 17.22 dB compared to RetinexNet’s 16.77 dB, and an SSIM of 0.5417 versus RetinexNet’s 0.4252. The hybrid optimization strategy of BOA and FA contributes significantly to this performance, offering a 12.3% improvement in PSNR and a 14.8% enhancement in SSIM over an unoptimized version of the pipeline.
In conclusion, BFORE represents a significant advancement in low-light image enhancement, combining innovative optimization techniques with a robust framework to achieve superior results. As the demand for high-quality imaging continues to grow across various applications, BFORE offers a promising solution to the challenges posed by low-light conditions.
Related AI Insights
- Multimodal LLMs Detect Seizure Movements: Pilot Study
- Automating RL Interfaces Using Large Language Models
- Top E Ink Tablet Recommended by Hundreds of Readers
- LLM Safety Flaws Revealed by Mathematical Encoding Attacks
- Perplexity’s AI Personal Computer Now on Mac
- DynaTab: Dynamic Feature Ordering for High-Dimensional Data
- Boost Reasoning Tasks with RAG Using Thinking Traces
- MEMSAD: Advanced Anomaly Detection for Memory Poisoning
- Bumble Ditches Swipe for AI-Powered Dating Assistant
- Deepfake Audio Detection with Self-Supervised Fusion
