Contrast enhancement-based no-reference image quality assessment method
Zongyuan Ge, Yu Ming Shen, Chaoliu Tong, Haikun Wei, Kanjian Zhang · 2025
No-Reference Image Quality Assessment (NR-IQA) aims to evaluate image quality without relying on a reference image. However, in current NR-IQA research, few methods assess image quality based on contrast. This paper proposes a simple yet effective NR-IQA method, leveraging the principle that high-contrast images are generally more similar to their contrast-enhanced versions, thereby facilitating quality evaluation. Specifically, the method first obtains a grayscale image using the maximum grayscale value method. Then, a histogram equalization technique is applied to generate a contrast-enhanced image. The structural similarity index (SSIM) between the original and enhanced images is extracted as the first image feature. Additionally, four more image features are obtained by computing the entropy and cross-entropy of the histograms of the original and enhanced images. Finally, an SVR regression module is employed to integrate these five image features and predict the image quality score. The proposed method has been validated on publicly available datasets, demonstrating its superiority and efficiency.