Comprehensive Evaluation Metric for Image Quality

Based on the contrast sensitivity from EF characteristics, specifically the spatial frequency response curve, the method performs two-dimensional multi-level wavelet decomposition on images. From the decomposed wavelet components, it extracts corresponding luminance, sharpness, and correlation metrics for each frequency band. The implementation typically involves using wavelet transform functions like wavedec2 in MATLAB or pywt.wavedec2 in Python's PyWavelets library to achieve multi-scale decomposition. Next, the geometric mean of these three metrics is calculated and then arithmetically averaged with the inner product of frequency band weighting coefficients. This weighted averaging process can be implemented through matrix multiplication operations where weighting coefficients are optimized based on human visual system characteristics. This approach yields a comprehensive metric for evaluating image quality that considers luminance, sharpness, and correlation across different frequency bands. Through such an evaluation metric, we can more comprehensively assess image quality by accounting for multi-scale spatial frequency characteristics and perceptual importance of different frequency components.