A very challenging goal for most image filtering systems consists in removing the noise without causing detail blur and loss of useful information. Thus, performance evaluation of new filters should carefully consider the measurement of these conflicting features. A new technique is presented in this paper to overcome the limitations of the sole peak signal-to-noise ratio (PSNR) and other metrics in evaluating the performance of greyscale image filtering systems. The proposed approach computes the PSNR component, namely peak signal-to-blur ratio (PSBR), that measures the unwanted distortion produced during noise removal. The approach reinforces the role of PSNR as quality index in image filtering. Indeed, the pair of PSNR and PSBR evaluations suffice to characterize the key behavior of a denoising system. Computer simulations dealing with images restored from Gaussian and impulse noise show that the new method is more effective than the available metrics.
