Chapter 8 - A comparative analysis of image restoration techniques

Abstract

Image restoration is the process of extracting clean images, i.e., beneficial information from the noisy and blurred input images. Image processing may be carried out for various image restoration purposes, such as astronomical imaging, medical imaging, digital media restoration, law enforcement, modular spectroscopy, etc. But the image is often corrupted by atmospheric noises, camera sensors, instrumentation noises, etc. This effect will decrease the accuracy of the image and harm the details it holds. A restoration process can restore a degraded image to its actual state. The image gets distorted due to camera motion of relative objects, sudden atmospheric turbulence, and the camera's poor focus. This chapter provides the workflow, methodologies, and comparative performance analysis of image restoration techniques. Some of the methods discussed in this chapter are filtering techniques, model selection criterion, hybrid kernel padding algorithm, fast Fourier transform, iterative denoising, and backward projection and boosting techniques. The performance of these methods is evaluated on several benchmark metrics, namely signal-to-noise ratio (SNR), mean square error (MSE) and peak signal-to-noise ratio (PSNR). After the detailed analysis, it is observed that among all the above evaluation parameters hybrid sparsity learning technique performs better than the other state-of-the-art techniques.