Image restoration using filters and image quality assessment using reduced reference metrics

This paper proposes restoration of an image using filters before an image quality assessment is performed. Image noising is being done to a quality image by addition of impulse noise in different variances. There are two methods of image restoration namely; adaptive median filter and improved median filter. These filters are used to perform impulse noise removal function. Then the quality assessment of the restored image with original image (reference image) is performed. Reduced-reference metrics is implemented, wherein it makes use of only limited information from the reference image. For implementing reduced reference metrics, 2D Fourier transform is used to obtain the phase and magnitude value. This paper presents better results in terms of the combination of phase and magnitude score denoted as Q_c. Based on Q_c score, the suitable filter is chosen for the particular value of variance of the impulse noise. The least Q_c score indicates the best quality image and highest score indicate the worst image quality. From the analysis with Q_c score, it was found that higher impulse noise image can be restored by means of improved median filter and the lower impulse noise image can be restored by means of adaptive median filter.