Contrast Enhancmement Metric
This project is an attempt to explore a human element not easily solved in the image processing communities. The problem statement is vague but important to address. What is a good image? More specifically, if a low contrast image is presented, at what level of enhancement is good enough for a human observer?
Help me by completing
the image comparison test! »
ISCAS 2012
Data [1]
Journal Code 2012
Under review
Color Ellipsoid Framework
Paper Under Review [7]
CEM
This work presents a new contrast enhancement metric (CEM) that is trained using several simple contrast measures and mean opinion scores obtained from human observations. Our goal is to train the algorithm to mimic a human when selecting an image with the best contrast between two images. For example, the algorithm will accept two images of the same scene with differing (unknown) contrast and will choose which of the two images is `better' according to what a human believes is `better'. See Figure 1 for an example of how to use the CEM.
-
Figure 1 - Block Diagram using CEM
Given an input image, the CEM will measure the performance of each contrast enhancement filter from the filter suite. The filter with the highest CEM value will be chose for that input image.
-
Figure 2 - Images Sorted by Perceptual Contrast
This image sent was sorted using the CEM by comparing to each image pair. From left to right is increase in perceptual contrast.
Journal Supplemental Materials
The following figures and tables are supplemental materials for the journal A No-Reference Perceptual Based Contrast Enhancement Metric for Ocean Scenes in Fog.
-
Figure 3 - Statistics of Test Subjects
Online Test Example and Demographics. The gender, age, education, expertise and enhancement usage statistics were captured for each observer and illustrated in each plot.
-
Figure 4 - Original Image Set
These are the original images used for the subjective tests along with 7 different enhanced versions.
-
Table 1 - List of methods applied to each scene
Each original image---which had fog or haze---was manipulated with 7 different methods in this table.
-
Table 2 - Individual Sub-metric results
The average error and standard deviation for each sub-metric.
-
Table 3 - Correlation Results
For the sake of space and clarity, we only show the results from the CEMhc and CEMh methods because they produced the highest averaged correlations. Each row is a correlation measurement for each scene. The last row is the average over all the scenes.
References
[1] He K, Sun J, Tang X: Single Image Haze Removal Using Dark Channel Prior. In CVPR 2009:1956-1963.[2] Chen Z, Abidi B, Page D, Abidi M: Gray-level grouping (GLG): an automatic method for optimized image contrast enhancement-Part II: the variations. In TIP 2006, 15(8):2303-14.
[3] Gonzalez RC, Woods RE: Digital Image Processing, 2nd edition 2001, :793.
[4] Gibson K, Vo D, Nguyen T: An Investigation of Dehazing E ects on Image and Video Coding. In TIP 2012, 21(2):662-673.
[5] Tarel JP, Hautiere N: Fast Visibility Restoration from a Single Color or Gray Level Image. In ICCV, Kyoto, Japan 2009:2201-2208. Webpage
[6] Gibson, K.; Nguyen, T.: A Perceptual Based Contrast Enhancement Metric using AdaBoost., ISCAS 2012, Seoul, Korea, May. PDF
[7] Gibson, K.; Nguyen, T.: An Analysis of Single Image Defogging Methods using a Color Ellipsoid Framework. In EURASIP, Under Review PDF