Presentation is loading. Please wait.

Presentation is loading. Please wait.

Introduction to Image Quality Assessment

Similar presentations


Presentation on theme: "Introduction to Image Quality Assessment"— Presentation transcript:

1 Introduction to Image Quality Assessment

2 Outline Applications Image Quality Assessment
Image Quality Assessment with Reference Image -Zhou Wang and Alan C. Bovik, ICASSP2002 Blind Image Quality Assessment -Xin Li, ICIP2002

3 Image Quality ?

4 Image Quality Assessment
Good Bad

5 Applications Image Acquisition Systems and Display Systems
Image Processing Systems and Algorithms Compression and Network

6 Image Quality Assessment
Mean Opinion Score Automatically Image Quality Evaluation

7 Mean-Squared Error and Peak Signal-to-Noise Ratio

8 Frequency-domain SNR

9 Frequency-domain SNR

10 Image Quality Assessment Methods
Image Quality Assessment with Reference Image Blind Image Quality Assessment

11 Image Quality Assessment Methods
Image Quality Assessment with Reference Image Blind Image Quality Assessment

12 Error Sensitivity Based Image Quality Measurement

13 Error Sensitivity Based Image Quality Measurement
alignment, luminance transformation, and color transformation

14 Error Sensitivity Based Image Quality Measurement
resulting in two sets of transformed signals for different channels

15 Error Sensitivity Based Image Quality Measurement
The errors between the two signals in each channel are calculated and weighted, usually by a Contrast Sensitivity Function (CSF).

16 Error Sensitivity Based Image Quality Measurement
The weighted error signals are adjusted by a visual masking effect model, which reflects the reduced visibility of errors presented on the background signal

17 Visual Masking Effect

18 Error Sensitivity Based Image Quality Measurement
Minkowski error pooling

19 Weaknesses of Error Sensitivity Based Methods
The reference signal is of perfect quality There exist visual channels in the HVS and the channel responses can be simulated by an appropriate set of channel transformations. CSF variance and intra-channel masking effects are the dominant factors that affect the HVS’s perception on each transformed coefficient in each channel OK!

20 Weaknesses of Error Sensitivity Based Methods
For a single coefficient in each channel, after CSF weighting and masking, the relationship between the magnitude of the error and the distortion perceived by the HVS can be modeled as a non-linear function. The interaction between channels is small enough to be ignored.

21 Weaknesses of Error Sensitivity Based Methods
The perceived image quality is determined in the early vision system. Higher level processes, such as feature extraction, pattern matching and cognitive understanding happening in the human brain, are less effective Active visual processes, such as the change of fixation points and the adaptive adjustment of spatial resolution because of attention, are less effective

22 Structure Distortion Based Image Quality Measurement
The main function of the human eyes is to extract structural information from the viewing field, and the human visual system is highly adapted for this purpose. Therefore, a measurement of structural distortion should be a good approximation of perceived image distortion

23 Structure Distortion Based Image Quality Measurement

24 Structure Distortion Based Image Quality Measurement
loss of correlation, mean distortion, and variance distortion

25 Experimental Results a: original b: 0.9372 c: 0.3891 d: 0.6494
f:

26 Image Quality Assessment Methods
Image Quality Assessment with Reference Image Blind Image Quality Assessment

27 Blind Image Quality Assessment
Human visual system usually does not need any reference to determine the subjective quality of a target image Distinction between fidelity and quality

28 Blind Image Quality Assessment
Edge Sharpness Level Random Noise Level Structured Noise Level

29 Edge Sharpness Level

30 Random Noise Level Impulse Noise Additive White Gaussian Noise

31 Structured Noise Level
Block Artifact Ringing Artifact

32 Blind Image Quality Assessment
Combination of different measurement is still a problem.

33 Conclusion Image Quality Assessment with Reference Image
Blind Image Quality Assessment


Download ppt "Introduction to Image Quality Assessment"

Similar presentations


Ads by Google