No-Reference Blur Assessment of Digital Pictures Based on Multifeature Classifiers

Abstract

In this paper, we address the problem of no-reference quality assessment for digital pictures corrupted with blur. We start with the generation of a large real image database containing pictures taken by human users in a variety of situations, and the conduction of subjective tests to generate the ground truth associated to those images. Based upon this ground truth, we select a number of high quality pictures and artificially degrade them with different intensities of simulated blur (gaussian and linear motion), totalling 6000 simulated blur images. We extensively evaluate the performance of state-of-the-art strategies for no-reference blur quantification in different blurring scenarios, and propose a paradigm for blur evaluation in which an effective method is pursued by combining several metrics and low-level image features. We test this paradigm by designing a no-reference quality assessment algorithm for blurred images which combines different metrics in a classifier based upon a neural network structure. Experimental results show that this leads to an improved performance that better reflects the images' ground truth. Finally, based upon the real image database, we show that the proposed method also outperforms other algorithms and metrics in realistic blur scenarios.

No-reference image and video quality assessment: a classification and review of recent approaches

“…Some methods, however, make use of HVS adaptation to the value of edge distortion to classify it as perceivable or not perceivable by a human subject. A paradigm for blur evaluation has been presented in [22] that is mainly composed of four methods of blur quantification, given in [23][24][25] and [26], which have been integrated by an artificial neural network (ANN) powered multifeature classifier. In the method given in [23], an image quality measurement method in terms of global blur has been proposed.…”

“…An improved version of [25] is found in [26] where HVS properties have been added to get weighted edge lengths. It is to be noted that none of these four reference methods quantify realistic blur situations, but Ciancio et al [22] have shown their method to be useable for measuring naturally occurring blur. Overall, [22] uses local phase coherence, mean brightness level, and variance of the HVS frequency response and contrast as additional inputs, together with the earlier mentioned four methods, to various ANN models designed for quality estimation.…”

An Improvement of Structural Similarity Index for Image Quality Assessment

“…In terms of gray scale, this means that the process of blurring leads into high intensity variations between adjacent pixels of a net image while only mild intensity variations between adjacent pixels are presented if the image is already blurred (Liu et al, 2008;Ciancio et al, 2011).…”

Image Blur Classification and Parameter Identification Using Two-stage Deep Belief Networks

“…For testing stage, these parameters in the network act like 'prior' information for the degraded images, which end up with better results compared to the top local denoising approaches. Another example is the blur extent metric developed by the multifeature classifier based on Neural Networks (NN) [9]. It has proved that the combined learned feature works better than individual handcrafted feature mostly.…”

“…Inspired by the practical blur type classification in [19,29] and the merits of learned descriptors in [9,13], we intend to design another patch-based blur type classification and parameters identification method to better solve the realistic blur analysis problem. Deep Belief Network (DBN) is chosen for accomplishing the feature extraction and final classification in this system.…”