Presented is a new no-reference blur index for still images that is based on the observation that it can be difficult to perceive between versions of an image blurred to different degrees. A 're-blurred' image is produced by intentionally blurring the test image. Local sample statistics are computed in the vicinity of detected edges of the original and reblurred images, respectively. These are differenced and normalised to construct a new blur index. Experimental results on four simulated blur databases and on the Real Blur Image Database show that the proposed method obtains high correlations with test subjective quality evaluations.Introduction: Image quality assessment plays a significant role in numerous image processing applications. Since images are usually created or processed for human visual consumption, subjective evaluation by human viewers is the ultimate measurement of image quality. However, subjective evaluation is usually too inconvenient, timeconsuming and expensive [1]. Therefore, developing objective image quality assessment (IQA) indices that can effectively assess the quality of an image is needed. Since the original image cannot be assumed to be available in many potential practical applications, the design of no-reference assessment algorithms is of great interest.Blur is one of the most common phenomena in images. In recent years, a variety of no-reference blur indexes have been proposed in the literature. Most of these seek to measure the overall sharpness of the image, the idea being that a lack of sharpness indicates probable blur. For example, in [2], the authors propose a new perceptual blur index that operates by blurring the test image, then comparing the amount of variation in the result to that in the original test image. In [3], an image sharpness index is proposed that is based on the notion of just noticeable blur (JNB). In it, image blocks are categorised as smooth or nonsmooth according to the density of edge pixels each contains. The estimated blur distortion for the image is then obtained by pooling the estimated blur distortion of each nonsmooth block, which is defined as a function of the widths of edge pixels and local contrasts. Recently, the authors of [4] proposed a new sharpness measure utilising local phase coherence (LPC) evaluated in the complex wavelet transform domain. A spatially varying LPC map is constructed for the input image. Pooling the LPC map by weighted averaging yields the global image sharpness index.Here we use an observation regarding blur perception [2] to create a novel blur estimation method which does not require the reference image. The new method uses sample statistics of the test image and a purposely blurred version of it to assess the blur quality.
Abstract-Median type filters take the main stream in suppressing impulse noise, and the Laplacian distribution assumption lays the basis for it. We however demonstrate in this paper that the Gaussian distribution assumption is more preferable than Laplacian distribution assumption in suppressing impulse noise, especially for high noise densities. This conclusion is supported by numerical experiments with different noise densities and filter models Index Terms-Median filter, Impulse noise, Salt and pepper noise, Gaussian distribution, Laplacian distribution.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.