Diffraction-based Ringing Suppression in Image Deblurring
Digital processing of images are essential parts of numerous applications ranging from biomedical image diagnosis to applications in machine vision and artificial intelligence. One of the common problems is image blurring effects due to degradation. Image blurring can inherently be caused by equipment (camera lens), motion of the target object or vibration of the image acquisition device while the image is being captured or other degradation processes. A number of image restorations algorithms have appeared in the literature using both blind and nonblind deconvolution techniques to restored a sharp latent image from its degraded blurred version. Most of these observed deblurred images exhibits an artifact (ringing effect) in the areas around the edges of the restored image. In this study, a method for decreasing these effects is proposed to improve the clarity of the objects in the image and reduce image obscurity. The method makes use of the realization that the ringing artifact also known as the Gibbs phenomenon are mostly caused by the distortion or loss of frequency contents of input image which is unavoidable as most of the deconvolution algorithms operations are performed in frequency domain with help of Fourier transforms. Area with high gradients magnitude (high frequency region) whose frequency contents are likely going to be distorted in the subsequent operations are estimated and modeled from the blurred input image using Gaussian low-pass filter. Then, concept of wave diffraction model using Fraunhofer diffraction equation is used to infused the estimated edge maps to compute the scaled and less distorted Fourier transform of the of the blurred image. Results from the experiment using MMIP ringing dataset show an improved result with suppressed ringing effect in both blind and nonblind deconvolution where the proposed method was used.