Discriminative Blur Detection Features Jianping Shi , Li Xu, Jiaya - - PowerPoint PPT Presentation

▶
discriminative blur detection features
SMART_READER_LITE
LIVE PREVIEW

Discriminative Blur Detection Features Jianping Shi , Li Xu, Jiaya - - PowerPoint PPT Presentation

Nov 08, 2022 307 likes •638 views

Discriminative Blur Detection Features Jianping Shi , Li Xu, Jiaya Jia CVPR, 2014 Image Blurriness Commonly occurred photo degradation Visual effect by photographers Important to analyze 10/20/2014 2 Image Blur Detection Problem

slide-1
SLIDE 1

Discriminative Blur Detection Features

Jianping Shi, Li Xu, Jiaya Jia

CVPR, 2014

slide-2
SLIDE 2

Image Blurriness

  • Commonly occurred photo degradation
  • Visual effect by photographers
  • Important to analyze

10/20/2014 2

slide-3
SLIDE 3

Image Blur Detection

  • Problem Definition

– Finding blur pixels for a given input image

  • Potential application

– Image segmentation, – Object detection, – Image quality assessment, – …

10/20/2014 3

slide-4
SLIDE 4

Previous Works

= * = *

  • De-convolution based approach

10/20/2014 4

slide-5
SLIDE 5

Previous Works

  • Deconvolution based approach
  • Explicit blur detection

– Gradient based: Levin 2007, Liu 2008 – Frequency based: Liu 2008, Chakrabarti 2010 – Matting based: Dai 2008, Dai 2009

10/20/2014 5

slide-6
SLIDE 6

Our Blur Features

  • Image Gradient Distribution
  • Spectra in Frequency Domain
  • Local Filters

10/20/2014 6

slide-7
SLIDE 7

Our Blur Features

  • Image Gradient Distribution

Properties:

  • 1. Peakedness
  • 2. Heavy-tailedness

10/20/2014 7

slide-8
SLIDE 8

Our Blur Features

  • Image Gradient Distribution

– Previous L0.8 norm

10/20/2014 8

slide-9
SLIDE 9

Our Blur Features

  • Image Gradient Distribution

– Peakedness Measure

  • Definition: Kurtosis
  • Proposition 1: Given the local blur model and kurtosis

measure, it is guaranteed to have 𝐿 𝐶𝑦 ≤ 𝐿(𝐽𝑦) and

𝐿 𝐶𝑧 ≤ 𝐿(𝐽𝑧).

10/20/2014 9

slide-10
SLIDE 10

Our Blur Features

  • Image Gradient Distribution

– Peakedness Measure

  • Kurtosis feature

10/20/2014 10

slide-11
SLIDE 11

Our Blur Features

  • Image Gradient Distribution

– Peakedness Measure

  • Kurtosis feature

10/20/2014 11

slide-12
SLIDE 12

Our Blur Features

  • Image Gradient Distribution

– Heavy-Tailedness Measure

  • Fit a Gaussian mixture model

with two components to gradient magnitude 𝛼𝐶 𝛼𝐶 ∼ 𝜌1𝐻(𝛼𝐶|𝜈1, 𝜏1)

+𝜌2𝐻(𝛼𝐶|𝜈2, 𝜏2)

10/20/2014 12

slide-13
SLIDE 13

Our Blur Features

  • Image Gradient Distribution

– Heavy-Tailedness Measure

  • Fit a Gaussian mixture model with two components to

gradient magnitude 𝛼𝐶

  • The heavy-tailedness feature is the larger variance

10/20/2014 13

slide-14
SLIDE 14

Our Blur Features

  • Spectra in Frequency Domain

– Average power spectrum 𝐾(𝜕)

  • It intuitively represents the strength of change
  • Blur attenuates high frequency components.
  • The power spectrum fall of faster for blur region

10/20/2014 14

slide-15
SLIDE 15

Our Blur Features

  • Spectra in Frequency Domain

– Average power spectrum 𝐾(𝜕)

– Proposition 2: Given a natural image patch 𝑦 and its Gaussian or box blurred version 𝑧 by PSF 𝑙, the fall-off speed of the average power spectrum on 𝑧 is several

  • rders faster than that of 𝑦. It is expressed as

10/20/2014 15

slide-16
SLIDE 16

Our Blur Features

  • Spectra in Frequency Domain

– Spectrum feature:

– Proposition 3: Given a natural image patch 𝑦, which is blurred by a PSF to form patch 𝑧, the cumulated average power spectrum for the blurred patch is smaller than that for the sharp patch, i.e.,

10/20/2014 16

slide-17
SLIDE 17

Our Blur Features

  • Spectra in Frequency Domain

10/20/2014 17

slide-18
SLIDE 18

Our Blur Features

  • Local Filters

– Data driven approach based on our labeled dataset – Linear discriminative analysis – Learned feature

10/20/2014 18

slide-19
SLIDE 19

Our Blur Features

  • Local Filters

10/20/2014 19

slide-20
SLIDE 20

Visualizing Features in 3D

10/20/2014 20

slide-21
SLIDE 21

Feature Covariance

10/20/2014 21

slide-22
SLIDE 22

Multi-Scale Perception

  • Scale ambiguous

10/20/2014 22

slide-23
SLIDE 23

Multi-Scale Perception

  • Fuse information in different scales
  • The input for each layer is the posterior of a naive

Bayesian classifier for the set of features.

10/20/2014 23

slide-24
SLIDE 24

Multi-Scale Perception

  • Fuse information in different scales

10/20/2014 24

slide-25
SLIDE 25

Blur Detection Dataset

10/20/2014 25

slide-26
SLIDE 26

Experiments

  • Visual comparisons
  • Quantitative comparisons
  • Applications enabled by blur detection

10/20/2014 26

slide-27
SLIDE 27

Visual Comparison

10/20/2014 27

slide-28
SLIDE 28

Quantitative Comparison

10/20/2014 28

slide-29
SLIDE 29

Applications Based on Blur Detection

  • Blur Segmentation and Deblurring

10/20/2014 29

slide-30
SLIDE 30

Blur Magnification

10/20/2014 30

slide-31
SLIDE 31

Failure Case

10/20/2014 31

slide-32
SLIDE 32

Conclusion

  • We have proposed several effective local blur

features

  • We have integrated the local blur features into

a multi-scale inference framework

  • Extensive experiments verified our method

10/20/2014 32