Image enhancement. Image enhancement belongs to image preprocessing

Transcription

1 Image enhancement Image enhancement belongs to image preprocessing methods. Objective o image enhancement process the image (e.g. contrast improvement, image sharpening, ) so that it is better suited or urther processing or analysis P. Strumiłło, M. Strzelecki

2 Image enhancement Image enhancement methods are based on subjective image quality criteria. o objective mathematical criteria are used or optimizing processing results. subjective perception

3 Image enhancemet methods Point processing Contrast enhancement Histogram modelling Image averaging Spatial iltering Linear ilters onlinear ilters Edge detection Zooming Image colouring Pseudo colouring False colouring

4 Image enhancement Brightness J M M i j ( i, j) Contrast C M M i j [ ( i, j) J ] M, image dimensions (i,j) gray level value at (i,j)

5 Image histogram J94, C9 Image brightness and contrast inluence image subjective quality perception J, C47 J9, C38

6 Image histogram Image : array[..m,..] o byte; Hist : array[..l-] o longint;... Hist:; or i: to M do or j: to do Inc( Hist[ Image[i, j] ] ); 5 5 5

7 Image histogram Source image Image histogram represents statistical distribution o image pixel brightnesses bright 5 dark

8 Linear gray scale transormation L- g m g(i,j) m (i,j) + d d g m ~ contrast d ~ brightness L- OUTPUT IMAGE SOURCE IMAGE POIT OPERATIO

9 MATLAB Demo image histogram

10 Histogram stretching g POIT OPERATIO? M I M A X L- g(i,j) (i,j)< MI L- MAX - ((i,j)- MI ), MI (i,j) MAX MI L- (i,j)> MAX

11 Histogram stretching - example MI, MAX

12 Grayscale inversion L- g L-

13 onlinear grayscale transormation L- ln(x) g(i,j) T( (i,j)) g sqrt(x) x exp(x) g γ correction L- OUTPUT IMAGE SOURCE IMAGE Grayscale normalization! POIT OPERATIO

14 onlinear grayscale transormation - example Source image

15 onlinear grayscale transormation - example Tx 5 5 Tsqrt(x) 8 6 4

16 onlinear grayscale transormation - example Te x 5 5 Tlog(x) 5 5

17 onlinear grayscale transormation - algorithm Example: square unction normalization: minimum value - -> maximum value > 55 ormalization coeicient: norm/55... or i: to M do or j: to do g[i,j]:round(sqr([i,j])*norm);...

18 onlinear grayscale transormation - algorithm Example: square unction (using look-up-table) lut : array[..55]o byte;... or k: to 55 do lut[k]:round(k*k*norm) or i: to M do or j: to do g[i,j]:lut[([i,j])];...

19 Enhacement o a telescope moon image Tb log(ax)

20 Image ehancement by image averaging Consider a noisy image: ( i, j) ( i, j) + η( i j) g, contaminated by additive noise η(i,j) o zero average an variance σ h that is not correlated to the image. We will show that ater averagings (acquisitions) o the noisy image g(i,j) the variance o noise component will be reduced to: ση σ η

21 Image ehancement by image averaging + + k k k k j i n j i j i n j i j i g ), ( ), ( )], ( ), ( [ ), ( WARIG! grayscale range

22 oise variance in the averaged image: ( ) { } η η η σ σ η η η + η + η + + η η η η σ E E E E E k k p k p k k k k k k k K Image ehancement by image averaging One can also show that the pick value o noise {n} is reduced by a actor o ater image averagings

23 Image averaging example Additive Gaussian noise 8 6 Addison-Wesley Microscope image o a cell

24 Cumulative histogram ensions image M L i M k hist i histc o array hists o array hist histogram cumulative histc histogram image hist i k dim,,...,, ]) / [ ( ] [ ; [..55] : [..55] :, single longint;

25 Cumulative histogram Histogram Cumulative histogram

26 Histogram equalization Histogram equalization aims at obtaining uniorm statistical distribution o image gray levels (uniorm probability density unction) By histogram equalization one gets: contrast enhancement image normalization

27 Histogram equalization p () p (g) g /(L-) L- g L- p ()hist[]/m p g (g)/(l-)

28 ] [ ) ( ] [ ) ( ) ( ) (,,, ) (, ) ( ) ( ) ( histc L M i hist L i p L g,...,lg L g i p L g L g u L du L dh h p du u p dh h p i i i g g g Histogram equalization

29 Histogram equalization Cumulative histogram.5 g Equalized histogram Histogram g ( L ) histc[ ]

30 Cumulative histogram - algorithm hist : array[..55] o longint; histc : array[..55] o single;... histc[]:hist[]; or k: to 55 do histc[k]:histc[k-]+hist[k];...

31 Histogram equalization

32 Histogram equalization - example

33 MATLAB Demo intensity adjustment

34 Correction o nonuniorm illumination