Degree Of Blur Research Articles

Adaptive multi-focus image fusion using a wavelet-based statistical sharpness measure

Multi-focus image fusion aims to combine a set of images that are captured from the same scene but with different focuses for producing another sharper image. The critical issue in the design of multi-focus image fusion algorithms is to evaluate the local content information of the input images. Motivated by the observation that the marginal distribution of the wavelet coefficients is different for images with different focus levels, a new statistical sharpness measure is proposed in this paper by exploiting the spreading of the wavelet coefficients distribution to measure the degree of the image's blur. Furthermore, the wavelet coefficients distribution is evaluated using a locally adaptive Laplacian mixture model. The proposed sharpness measure is then exploited to perform adaptive image fusion in wavelet domain. Extensive experiments are conducted using three sets of test images under three objective metrics to demonstrate the superior performance of the proposed approach.

基于背景先验点的虚焦抠图方法

本文针对图像抠图问题中背景虚焦时的情况，将一些模糊的背景预划分成先验点，使之成为绝对背景，进而避免了与前景颜色相近背景的干扰，使得抠图结果更为理想。首先，采用边缘检测器进行边缘检测，进而对每个边缘点进行分析，根据其对二阶算子的响应值与理想响应值之间的拟合误差，以及它与已知前景之间的颜色差异，判断其是否为被模糊的背景边缘点；然后再计算它沿梯度正负方向两侧的像素与理想化边缘之间的拟合误差，根据阈值可以将两侧的若干点扩展成背景先验点集，即将它们划分为绝对背景点。最后利用与trimap扩展相结合的抠图方式，得到最终的透明度图像。实验结果表明，本文提出的背景先验点与抠图算法，可以得到相对于以往的抠图算法更好的结果。 In this paper, we present a matting method based on focused foreground and blurred background. We classify some pixels as priors to prevent the similarities between foreground and background colors. Firstly, we design a three-channel edge detector to roughly predict some edge pixels. Secondly, for each edge pixel, we estimate its blur degree by fitting an ideal second derivative filter response on the actual one along the gradient direction, and comparing its color and known foreground one. Thirdly, if this pixel is classified into a blurred background edge, we extend it along the gradient direction as blur priors according to the fitting errors to an ideal blurred edge. Finally, with the back-ground blur priors, we run a general matting algorithm along with a trimap expansion method. The experimental results show that our background blur priors could generate much more precise alpha results than the state-of-art algorithms.

A unified approach to autofocus and alignment for pattern localization using hybrid weighted Hausdorff distance

Pattern localization is a fundamental task in machine vision, and autofocus is a requirement for any automated inspection system by allowing greater variation in the distance from the camera to the object being imaged. In this paper, we propose a unified approach to simultaneous autofocus and alignment for pattern localization by extending the idea of image reference approach. Under the least trimmed squares (LTS) scheme, the proposed hybrid weighted Hausdorff distance (HWHD) is a robust similarity metric that combines the Hausdorff distance (HD) with the edge-amplitude normalized gradient (EANG) matching. The EANG is designed to characterize the different degrees of blur at the edge points for focus cues, immune to illumination variations between the reference and the target image. We experimentally illustrate its performance on simulated as well as real data.

주파수 도메인 정보를 이용한 영상의 Sharpness 평가 방법

본 논문에서는 영상의 선명도를 영상의 주파수 도메인 정보를 이용하여 측정하는 새로운 무기준법 화질 평가 방법을 제안한다. 기존에, 영상에 대한 선명도는 일반적으로 영상의 픽셀 값을 이용하여 측정되었다. 제안된 방법은 기존 방법과 달리 영상에 대한 선명도를 주파수 도메인 정보를 이용하여 측정하였다. 주파수 도메인에서 선명도를 평가하기 위하여 주어진 영상은 가우시안 저주파 필터를 사용하여 열화 되고, 열화 영상과 주어진 영상의 주파수 영역 계수를 사용하여 새로운 선명도 평가 함수를 정의하였다. 제안된 방법의 성능 검증은 TID2008 화질 평가 데이터베이스를 사용하여 이루어졌다. 기존 무기준법 영상 선명도 평가 방법과 비교하였을 때, 제안된 선명도 평가 방법은 주관적 화질 점수와 보다 높은 상관도를 보였다. In this paper, a new no-reference sharpness measure using frequency domain coefficients is proposed. Although most existing sharpness measures used pixel intensity to compute the blur degree, the proposed sharpness measure computes the sharpness using frequency coefficients. To assess the perceived sharpness of a given image, the image is re-blurred by a Gaussian low pass filter and a new quality measure function was defined using the frequency domain coefficients of the given image and the re-blurred image. To evaluate the proposed algorithms, TID2008 quality assessment database was used. Experimental results show that the proposed quality assessment method showed high correlation with the subjective scores.

Determination of Stripe Edge Blurring for Depth Sensing

Estimation of the blurring effect is very important for many imaging systems. This letter reports an idea to efficiently and robustly compute the blurring parameter on certain stripe edges. Two formulas are found to determine the degree of imaging blur only by calculating the area sizes under the corresponding profile curves, without the need for deconvolution or transformation over the image. The method can be applied to many applications such as vision sensing of scene depth. A 3-D vision system is taken as an implementation instance.

A behavioural method for evaluating depth of focus

Abstract Purpose To develop a technique based on visual reaction times that can evaluate the depth of focus in uncorrected vision or when using different modalities of simultaneous vision correction for presbyopia. Reaction Times (RTs) are known to increase in a predictable way as stimuli become less visible, i.e. as contrast decreases. Contrast reduction, produced by different amounts of defocus blur, was calculated by increased RT. Methods RTs were measured to the onset of a sinusoidal grating with a spatial frequency of 4 c/deg and a contrast of 10%, displayed on a Sony F‐520 CRT monitor (mean luminance 30 cd/m2), using a VSG2/5 stimulus generator card. All measurements were performed monocularly (dominant eye) and binocularly at 3m viewing distance, with best spectacle sphero‐cylindrical correction for distance and natural pupils. RTs were measured for different degrees of blur, induced using positive powered lenses up to +4.50 D. Results RT increased with defocus blur, with the effect being more pronounced for monocular compared to binocular viewing conditions. Depth of acceptable perceptual blur (depth of focus) was calculated from the second order polynomials, which fitted the RT vs. defocus blur functions. More specifically, the width of the polynomial fit was estimated for a specific RT. Regardless of the criterion chosen, the depth of acceptable blur was greater under binocular to monocular viewing. Conclusion The technique described here offers a behavioural measure of the depth of focus, and allows evaluation under both monocular / binocular viewing conditions.

Retinal blur and the perception of egocentric distance

A central function of vision is determining the layout and size of objects in the visual field, both of which require knowledge of egocentric distance (the distance of an object from the observer). A wide range of visual cues can reliably signal relative depth relations among objects, but retinal signals directly specifying distance to an object are limited. A potential source of distance information is the pattern of blurring on the retina, since nearer fixation generally produces larger gradients of blur on the extra-foveal retina. While prior studies implicated blur as only a qualitative cue for relative depth ordering, we find that retinal blur gradients can act as a quantitative cue to distance. Surfaces depicted with blur gradients were judged as significantly closer than those without, with the size of the effect modulated by the degree of blur, as well as the availability of other extra-retinal cues to distance. Blur gradients produced substantial changes in perceived distance regardless of relative depth relations of the surfaces indicated by other cues, suggesting that it operates as a robust cue to distance, consistent with the empirical relationship between blur and fixation distance.

ADAPTIVE NONLINEAR IMAGE ENHANCEMENT OF GAUSSIAN DEGRADED IMAGES

An adaptive technique for nonlinear image enhancement using Gabor filters is proposed. A set of Gabor filters are employed to extract high-pass components from the blurred image and these components are then nonlinearly processed before adding back to the input image for enhancement. Further, we propose a novel method for fast blur estimation and we establish an empirical relationship between the estimated blur and the optimal Gabor filter parameters, resulting in an enhancement system that is adaptive to the degree of blur in the input image. Extensive evaluation, including both PSNR-based objective evaluation and subjective psychophysical tests, confirms the advantages of the proposed approach over existing state-of-the-art methods. This enhancement approach is especially targeted at digital television applications where image blur is present due to various reasons like compression and resolution up-conversion.

Seeing edge blur: receptive fields as multi-scale neural templates

Edge blur is an important perceptual cue, but how does the visual system encode the degree of blur at edges? Blur could be measured by the width of the luminance gradient profile, peak ^ trough separation in the 2nd derivative profile, or the ratio of 1st-to-3rd derivative magnitudes. In template models, the system would store a set of templates of different sizes and find which one best fits the `signature' of the edge. The signature could be the luminance profile itself, or one of its spatial derivatives. I tested these possibilities in blur-matching experiments. In a 2AFC staircase procedure, observers adjusted the blur of Gaussian edges (30% contrast) to match the perceived blur of various non-Gaussian test edges. In experiment 1, test stimuli were mixtures of 2 Gaussian edges (eg 10 and 30 min of arc blur) at the same location, while in experiment 2, test stimuli were formed from a blurred edge sharpened to different extents by a compressive transformation. Predictions of the various models were tested against the blur-matching data, but only one model was strongly supported. This was the template model, in which the input signature is the 2nd derivative of the luminance profile, and the templates are applied to this signature at the zero-crossings. The templates are Gaussian derivative receptive fields that covary in width and length to form a self-similar set (ie same shape, different sizes). This naturally predicts that shorter edges should look sharper. As edge length gets shorter, responses of longer templates drop more than shorter ones, and so the response distribution shifts towards shorter (smaller) templates, signalling a sharper edge. The data confirmed this, including the scale-invariance implied by self-similarity, and a good fit was obtained from templates with a length-to-width ratio of about 1. The simultaneous analysis of edge blur and edge location may offer a new solution to the multiscale problem in edge detection.

The response dynamics of rabbit retinal ganglion cells to simulated blur

Retinal ganglion cells (RGCs) are highly sensitive to changes in contrast, which is crucial for the detection of edges in a visual scene. However, in the natural environment, edges do not just vary in contrast, but edges also vary in the degree of blur, which can be caused by distance from the plane of fixation, motion, and shadows. Hence, blur is as much a characteristic of an edge as luminance contrast, yet its effects on the responses of RGCs are largely unexplored.We examined the responses of rabbit RGCs to sharp edges varying by contrast and also to high-contrast edges varying by blur. The width of the blur profile ranged from 0.73 to 13.05 deg of visual angle. For most RGCs, blurring a high-contrast edge produced the same pattern of reduction of response strength and increase in latency as decreasing the contrast of a sharp edge. In support of this, we found a significant correlation between the amount of blur required to reduce the response by 50% and the size of the receptive fields, suggesting that blur may operate by reducing the range of luminance values within the receptive field. These RGCs cannot individually encode for blur, and blur could only be estimated by comparing the responses of populations of neurons with different receptive field sizes. However, some RGCs showed a different pattern of changes in latency and magnitude with changes in contrast and blur; these neurons could encode blur directly.We also tested whether the response of a RGC to a blurred edge was linear, that is, whether the response of a neuron to a sharp edge was equal to the response to a blurred edge plus the response to the missing spatial components that were the difference between a sharp and blurred edge. Brisk-sustained cells were more linear; however, brisk-transient cells exhibited both linear and nonlinear behavior.

The response dynamics of primate visual cortical neurons to simulated optical blur

Neurons in visual cortical area V1 typically respond well to lines or edges of specific orientations. There have been many studies investigating how the responses of these neurons to an oriented edge are affected by changes in luminance contrast. However, in natural images, edges vary not only in contrast but also in the degree of blur, both because of changes in focus and also because shadows are not sharp. The effect of blur on the response dynamics of visual cortical neurons has not been explored. We presented luminance-defined single edges in the receptive fields of parafoveal (1-6 deg eccentric) V1 neurons of two macaque monkeys trained to fixate a spot of light. We varied the width of the blurred region of the edge stimuli up to 0.36 deg of visual angle. Even though the neurons responded robustly to stimuli that only contained high spatial frequencies and 0.36 deg is much larger than the limits of acuity at this eccentricity, changing the degree of blur had minimal effect on the responses of these neurons to the edge. Primates need to measure blur at the fovea to evaluate image quality and control accommodation, but this might only involve a specialist subpopulation of neurons. If visual cortical neurons in general responded differently to sharp and blurred stimuli, then this could provide a cue for form perception, for example, by helping to disambiguate the luminance edges created by real objects from those created by shadows. On the other hand, it might be important to avoid the distraction of changing blur as objects move in and out of the plane of fixation. Our results support the latter hypothesis: the responses of parafoveal V1 neurons are largely unaffected by changes in blur over a wide range.

Holographic Data Storage Channel Model With Intensity Factor

We consider the holographic data storage channel model which includes intensity factor. We define the intensity factor as the degree of the diminishing phenomenon of the read page and model the diminishing phenomenon using a simple page masking filter. Thus, the page masking filter coefficients are determined by the intensity factor. We discuss the performance of the channel model employing one-dimensional (1-D) PRML with two-dimensional (2-D) PR equalizer when the degree of blur is 1.7.

Amblyopia masks the scale invariance of normal central vision

In normal vision, detecting a kink (a change in orientation) in a line is scale invariant: it depends solely on the length/width ratio of the line (D. Whitaker, D. M. Levi, & G. J. Kennedy, 2008). Here we measure detection of a change in the orientation of lines of different length and blur and show that strabismic amblyopia is qualitatively different from normal foveal vision, in that: 1) stimulus blur has little effect on performance in the amblyopic eye, and 2) integration of orientation information follows a different rule. In normal foveal vision, performance improves in proportion to the square root of the ratio of line length to blur (L:B). In strabismic amblyopia improvement is proportional to line length. Our results are consistent with a substantial degree of internal neural blur in first-order cortical filters. This internal blur results in a loss of scale invariance in the amblyopic visual system. Peripheral vision also shows much less effect of stimulus blur and a failure of scale invariance, similar to the central vision of strabismic amblyopes. Our results suggest that both peripheral vision and strabismic amblyopia share a common bottleneck in having a truncated range of spatial mechanisms--a range that becomes more restricted with increasing eccentricity and depth of amblyopia.

Fluency misattribution and visual hindsight bias

We tested a fluency-misattribution theory of visual hindsight bias, and examined how perceptual and conceptual fluency contribute to the bias. In Experiment 1a observers identified celebrity faces that began blurred and then clarified (Forward baseline), or indicated when faces that began clear and then blurred were no longer recognisable (Backward baseline). In surprise memory tests that followed, observers adjusted the degree of blur of each face to match what the faces looked like when identified in the corresponding baseline condition. Hindsight bias was observed in the Forward condition: During the memory test observers adjusted the faces to be more blurry than when originally identified during baseline. These same observers did not show hindsight bias in the Backward condition: Here, they adjusted faces to the exact blur level at which they identified the faces during baseline. Experiment 1b tested a combined condition in which faces were viewed in a Forward progression at baseline but in a Backward progression at test. Hindsight bias was observed in this condition but was significantly less than the bias observed in the Experiment 1a Forward condition. Experiments 1a and 1b provide support for the fluency-misattribution account of visual hindsight bias: When observers are made aware of why fluency has been enhanced (i.e., in the Backward condition) they are better able to discount it, and as a result show reduced or no hindsight bias. In Experiment 2, observers viewed faces in a Forward progression at baseline and then in a Forward upright or inverted progression at test. Hindsight bias occurred in both conditions, but was greater for upright than inverted faces. We conclude that both conceptual and perceptual fluency contribute to visual hindsight bias.

Reconstructing arbitrarily focused images from two differently focused images using linear filters

We present a novel filtering method for reconstructing an all-in-focus image or an arbitrarily focused image from two images that are focused differently. The method can arbitrarily manipulate the degree of blur of the objects using linear filters without segmentation. The filters are uniquely determined from a linear imaging model in the Fourier domain. An effective and accurate blur estimation method is developed. The simulation results show that the accuracy and computational time of the proposed method are improved compared with the previous iterative method and that the effects of blur estimation error on the quality of the reconstructed image are very small. The method performs well for real images acquired without visible artifacts.

Target spatial frequency determines the response to conflicting defocus- and convergence-driven accommodative stimuli

Asthenopia, or visual fatigue, is a frequent complaint from observers of stereoscopic three-dimensional displays. It has been proposed that asthenopia is a consequence of anomalous oculomotor responses generated by conflict between accommodative and convergence stimuli. The hypothesis was examined by measuring accommodation and convergence continuously with a Shin-Nippon SRW5000 infrared autorefractor and a limbus tracking device. Subjects viewed a high contrast Maltese Cross target at three levels of Gaussian filter target blur under conditions of relatively low- and high-conflict between accommodation and convergence stimuli, the latter inducing the sensation of stereopsis. Under the low-conflict conditions accommodation was stable, but convergence-driven accommodation was dominant when the target was extremely blurred. Under the high-conflict conditions the role of convergence-driven accommodation increased systematically with the degree of target blur. It is proposed that defocus-driven accommodation becomes weak when the target comprises low spatial frequency components. Large accommodative overshoots to step stimuli that are not blurred or only mildly blurred were consistently observed and are attributed to the initial accommodative response being convergence-driven. Whereas the possibility that high-conflict conditions are a cause of asthenopia has been previously reported, this is the first evidence that they specifically affect accommodative responses while viewing stereoscopic displays.

19.3: Correlation between Perceived Motion Blur and MPRT Measurement

The MPRT has been studied as the method of measuring motion blur on LCD. Subjective evaluation was conducted using the method of adjustment for investigating perceived motion blur. As a result, a strong correlativity was found between the perceived motion blur and the measurements by MPRT. So the MPRT can indicate very well the degree of motion blur perceived by humans.

Redundancy of stereoscopic images: Experimental evaluation

With the recent advancement in visualization devices over the last years, we are seeing a growing market for stereoscopic content. In order to convey 3D content by means of stereoscopic displays, one needs to transmit and display at least 2 points of view of the video content. This has profound implications on the resources required to transmit the content, as well as demands on the complexity of the visualization system. It is known that stereoscopic images are redundant which may prove useful for compression and may have positive effect on the construction of the visualization device. In this paper we describe an experimental evaluation of data redundancy in color stereoscopic images. In the experiments with computer generated and real life test stereo images, several observers visually tested the stereopsis threshold and accuracy of parallax measurement in anaglyphs and stereograms as functions of the blur degree of one of two stereo images. In addition, we tested the color saturation threshold in one of two stereo images for which full color 3D perception with no visible color degradations was maintained. The experiments support a theoretical estimate that one has to add, to data required to reproduce one of two stereoscopic images, only several percents of that amount of data in order to achieve stereoscopic perception.

Edges and bars: where do people see features in 1-D images?

There have been two main approaches to feature detection in human and computer vision––based either on the luminance distribution and its spatial derivatives, or on the spatial distribution of local contrast energy. Thus, bars and edges might arise from peaks of luminance and luminance gradient respectively, or bars and edges might be found at peaks of local energy, where local phases are aligned across spatial frequency. This basic issue of definition is important because it guides more detailed models and interpretations of early vision. Which approach better describes the perceived positions of features in images? We used the class of 1-D images defined by Morrone and Burr in which the amplitude spectrum is that of a (partially blurred) square-wave and all Fourier components have a common phase. Observers used a cursor to mark where bars and edges were seen for different test phases (Experiment 1) or judged the spatial alignment of contours that had different phases (e.g. 0° and 45°; Experiment 2). The feature positions defined by both tasks shifted systematically to the left or right according to the sign of the phase offset, increasing with the degree of blur. These shifts were well predicted by the location of luminance peaks (bars) and gradient peaks (edges), but not by energy peaks which (by design) predicted no shift at all. These results encourage models based on a Gaussian-derivative framework, but do not support the idea that human vision uses points of phase alignment to find local, first-order features. Nevertheless, we argue that both approaches are presently incomplete and a better understanding of early vision may combine insights from both.

Effects of imaging system blur on measurements of flow scalars and scalar gradients

Planar imaging of flow scalars is widely used in fluid mechanics, but the effects of imaging system blur on the measured scalar and its gradients are often inadequately quantified. Here, we present a 1-D analytical study that uses simplified models of the scalar profiles and imaging system blur to estimate the measurement errors caused by finite resolution. One objective of this paper is to give the experimentalist a methodology for quantitatively assessing the impact of imaging system blur on the accuracy of scalar measurements. The scalar profiles are modeled as either error or Gaussian functions, and the imaging system resolution is cast in terms of the line-spread function (LSF), which is modeled as Gaussian. The analysis gives the errors induced in the scalar structure thickness, gradient, and dissipation, for varying degrees of blur, the latter of which is quantified by σ, the standard deviation of the Gaussian LSF. The results show that, to keep errors in the peak scalar gradients and dissipation to less than 10%, the 20%-width of the scalar structures should be at least 7.5σ. Typical flow imaging experiments require fast (i.e., low f/#) optics that may suffer from significant blur and, therefore, this requirement may be difficult to meet in many applications. It is also shown that the resolution requirements for measuring the dissipation are more restrictive than for structure thicknesses. Further simulations were made to assess the effects of having clustered, or closely spaced, dissipation structures. Compared to the single structure results, there is a less severe resolution requirement to obtain scalar structure length scales, but a more severe requirement on the scalar gradient and dissipation.