A new approach for image quality assessment: FSM (Frequency Similarity Method)

Abstract

In this paper, we introduce a new approach for image quality assessment. The proposed method estimates similarity of frequency between a distorted image and a reference image using 8×8 block based DCT. Different from existing method such as SSIM (Structural Similarity Index Method) and PSNR, it does not use any data such as intensity, mean, variance, and covariance on the pixel domain. Instead, it simply uses only frequency components. As an experimental result, our new method achieves 90 percent performance to SSIM. However, the computational complexity of our proposed metric is simpler than SSIM. Since it uses a fixed number of coefficients for matrix multiplication, it can be easily implemented on hardware. Besides, our metric can use the DCT block which is already built in video system. If the original DCT block in video system is used for FSM, additional hardware resource is not needed to measure image quality.