Image quality measurement in the frequency domain

In this paper a novel frequency domain technique estimating the significance of blocking artifacts in images coded using the Discrete Cosine Transform is proposed. First a gradient image is computed which is then split into blocks. For each block, the blocking artifacts in the horizontal and vertical directions are detected by comparing ratios obtained from harmonics in each block with the corresponding ratios in the same block of the reference image. An image quality metric based on the significance of the differences of the ratios between the test picture and the reference picture for each block is introduced.