Texture,pattern and pattern classes

Texture,pattern and pattern classes

1. Texture,pattern and pattern classes M.Rajshree II-MSC (IT) Nadar saraswathi college of arts and science

2. Texture An important approach to region description is to quantify its texture content The three principal approaches used in image processing to describe texture of a region are Statistical Structural Spectral

3. Statistical Statistical approaches yield characterization of texture as smooth,coarse,grainy One of the simplest approaches for describing texture is to use statistical moments of gray level histogram of image or region Let z be a random variable denoting gray level and let p(zi),i=0,1,2…L-1 be corresponding histogram where L is number of distinct gray level

4. Contd… The mean just tell us average gray level of each region and it useful only as rough idea of intensity not really texture Fifth higher moments are not so easily related to histogram shape but they do provide further quantitative discrimination of texture content P’s have values in range [0,1] and their sum equal 1 measure U is maximum for an image which all gray level are equal

5. Structural Structural techniques deal with arrangement of image primitive such as description of texture based on regularly spaced parallel line We have a rule of form S aS which indicate that symbol S may be rewritten as aS The basic idea in foregoing discussion is that simple texture primitive can be used to form more complex texture pattern

6. Spectral Spectral technique are based on properties of fourier spectrum and used primarily to detect global in an image by high energy,narrow peaks in the spectrum Fourier spectrum ideally suited for describing directionality of periodic 2D pattern in an image Descriptor typically used for purpose are location of highest value mean and varience of both amplitude and axial variation

7. Pattern and pattern classes A pattern is an arrangement of descriptors A pattern class is a family of pattern that share some common properties Pattern class are denoted w1,w2,…wn where N is number of classes Pattern recognition by machine involve techniques for assigning pattern to their respective classes

8. There are three common pattern arrangemen t used Numeric vectors x=x1 x2…xn Strings and tress x=abababa… Recognition based on decision theoretic method These method are based on use of decision functions let x=(x1,x2,..xn) represent an n-dimensional pattern vector

9. Matching For N known pattern classes w1,w2,…wn idea here is to find N decision function d1(x),d2(x),…dn(x) Recognition techniques based on matching represent each class by a prototype pattern vector set of pattern of known classes is called testing set An unknown pattern is assigned to class to which it is closest in term of predefined metric

10. Wavelet based face recognition application Each face image in traning set is transformed to the wavelet domain to extract its pattern vector The choice of an appropriate varies depending on operational circumstance of face regonition application The decomposition level is predetermined based on efficiency and accuracy requirement and size of the face image

11. Contd… In the recognition stage a minimum distance classification method was used to classify the unknown face images Enrolment stage wavelet transform is applied on each training image so that a set W(f) of multi resolution decomposed images result A minimum distance classifier is used to classify input face image When a probe face image is introduced to system it decomposed by wavelet tranform certain subband chosen represent feature vector of probe image

12. Thank you