2. 1
Why another contrast enhancement technique?
โข Standard HE results in excessive contrast enhancement.
โข Lack of control of enhancement level.
โข Distortion in local details.
Proposed Technique for contrast enhancement
Histogram
Modification
Contrast Limited
Adaptive Histogram
Equalization
Input
Image
Output
Image
Database and Evaluation method used
โข Mammographic Image Analysis Society (MIAS) database.
โข Each grayscale image is 1024 x 1024 pixels.
โข Enhancement Measure (EME) as evaluation method.
3. 2
Histogram
0 7 3 2 3
0 0 0 6 7
7 7 2 2 0
1 1 0 4 1
0 0 7 4 1
Image matrixImage
0 1 2 3 4 5 6 7
Number of pixels of
intensity r
Histogram plots the number of pixels for each intensity value.
h(r)
Intensity values
4. 3 Histogram Modification
โข Increase the potentiality of image contrast enhancement.
โข Resultant image to be more relevant to input image.
โข Multi-objective optimization problem
Solution of the above problem
- modified histogram
- original (input) histogram
- uniform histogram
- enhancement parameter
0 < < โ
5. 4
Histogram Modification
Fig. mapping for enhanced images with
ัฑ = [0 1] โ 10 values.
Fig. mapping for enhanced images with
ัฑ = 0, 0.1, 0.5 and 1from top respectively
Mapping function used
๐ ๐ = ( 2 ๐ฟ
โ 1 ) เท
๐=0
๐
๐ ๐ ๐คโ๐๐๐ ๐ ๐ ๐๐ ๐๐๐ ๐๐ ๐๐๐๐๐๐๐๐ โ๐๐ ๐ก๐๐๐๐๐
6. 5
Contrast Limited Adaptive Histogram Equalization(CLAHE)
โข Unlike HE, CLAHE operates on small regions called tiles in the
images.
โข Contrast adjust according to their neighborhood pixels.
โข Distribution parameter is used to define the shape of the
histogram.
โข Original histogram is clipped and clipped pixels are redistributed
to each gray level.
โข Selecting appropriate clipping level, undesired noise amplification
can be reduced.
โข Different from AHE in terms of its contrasting limit.
โข Neighboring tiles are combined using bilinear interpolation to
eliminate artificially induced boundaries.
โข Clipping factor used in this experiment is 0.02
7. Obtain all
input values
Divide image into
contextual
regions/tiles
Enhancement
process applied
over tiles
Generate gray
level mapping
Apply mapping to
the image
Input
Image
Output
Image
CLAHE Algorithm
6
8. 7
Results and Performance Measure
โข The proposed methodology generates strong contrast enhancement.
โข All the local information of original image are preserved.
โข For this experiment ัฑ = 0.8 and clipping factor = 0.02
โข Performance is measured using Enhancement measure(EME).
โข Ideally should not be too high or too low.
Fig. EME for MIAS database