Deep Multi-Scale Feature Learning for Defocus Blur Estimation

Karaali, Ali; Harte, Naomi; Jung, Cláudio Rosito

doi:10.1109/tip.2021.3139243

Cited by 17 publications

(7 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We compare the quality scores of the deblurred images Î with the results of other two spatially varying defocus blur removal methods: a combination of the deconvolution methods proposed in [22] and [23], which is also used in [9,11,12,14], and the method proposed in [24]. Additionally, we compare the proposed method with a very recent blind approach [8] that estimates both the blur map and the deblurred Table 1 shows the SSIM and PSNR quality scores of the proposed approach and other competitive deblurring methods.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

SVBR-Net: A Non-Blind Spatially Varying Defocus Blur Removal Network

Karaali

Jung

2022

2022 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

Defocus blur is a physical consequence of the optical sensors used in most cameras. Although it can be used as a photographic style, it is commonly viewed as an image degradation modeled as the convolution of a sharp image with a spatially-varying blur kernel. Motivated by the advance of blur estimation methods in the past years, we propose a non-blind approach for image deblurring that can deal with spatially-varying kernels. We introduce two encoder-decoder sub-networks that are fed with the blurry image and the estimated blur map, respectively, and produce as output the deblurred (deconvolved) image. Each sub-network presents several skip connections that allow data propagation from layers spread apart, and also inter-subnetwork skip connections that ease the communication between the modules. The network is trained with synthetically blur kernels that are augmented to emulate blur maps produced by existing blur estimation methods, and our experimental results show that our method works well when combined with a variety of blur estimation methods. Our code will be available at https: //github.com/alikaraali/SVBR_Net_icip.

show abstract

Section: Resultsmentioning

confidence: 99%

“…In a highly related problem, defocus blur estimation approaches aim to estimate the spatially-varying blur kernel h(x, y) from a single blurry image [9][10][11][12][13][14], with increasingly better results. Such advance might leverage the development of non-blind deblurring methods, which assume that both I b and h are known.…”

Section: Introductionmentioning

confidence: 99%

SVBR-Net: A Non-Blind Spatially Varying Defocus Blur Removal Network

Karaali

Jung

2022

2022 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

show abstract

“…Figure 16 shows the original image (a) and the results obtained by various sharpening methods: fixed UM (b), BUM (c), and NLBM (d). Our method FIGURE 14 Intermediate blur maps extracted from [46] (middle row) and [47] (bottom row) using out-of-focus and motion blur images provides a satisfactory sharpening effect everywhere (different parts of the road), better than the one by BUM, and without incurring noise amplification (observe the sky region) as fixed UM does.…”

Section: Figurementioning

confidence: 92%

“…In order to detect blur in small regions, low level and high level features are combined for the blur map in contrast to using a local-global multicolumn architecture in [31]. While most deep learning based blur map generation aims towards a binary blur map, there are some deblurring CNNs which estimate the blur maps [46,47] as a prerequisite to generate deblurred images. These maps are further used for deblurring purposes.…”

Section: State Of the Art In Blur Analysismentioning

confidence: 99%

“…Indeed, the blur map that our method provides enables the user to enhance the input image also coping with the continuous-plane blur case mentioned in Section 2.1. The multilevel map activates the sharpening FIGURE 12 Blur maps derived with the method in [30] (middle row), and our method (bottom row) using sample images from CUHK dataset FIGURE 13 Intermediate blur maps extracted from [46] (middle row) and [47] (bottom row) using 2 plane and continuous plane images. While the BENET maps are smoother, the DIDANET maps show distinct edge information operator progressively, enhancing also partially blurred details up to the level in which no noise amplification is perceived in the smooth areas of the image.…”

Section: Figurementioning

confidence: 99%

See 1 more Smart Citation

Nonlinear kernel based feature maps for blur‐sensitive unsharp masking of JPEG images

Bhattacharya

Marsi

Ramponi

2022

IET Image Processing

View full text Add to dashboard Cite

In this paper, a method for estimating the blur regions of an image is first proposed, resorting to a mixture of linear and nonlinear convolutional kernels. The blur map obtained is then utilized to enhance images such that the enhancement strength is an inverse function of the amount of measured blur. The blur map can also be used for tasks such as attention‐based object classification, low light image enhancement, and more. A CNN architecture is trained with nonlinear upsampling layers using a standard blur detection benchmark dataset, with the help of blur target maps. Further, it is proposed to use the same architecture to build maps of areas affected by the typical JPEG artifacts, ringing and blockiness. The blur map and the artifact map pair permit to build an activation map for the enhancement of a (possibly JPEG compressed) image. Extensive experiments on standard test images verify the quality of the maps obtained using the algorithm and their effectiveness in locally controlling the enhancement, for superior perceptual quality. Last but not least, the computation time for generating these maps is much lower than the one of other comparable algorithms.

show abstract