Texture Analysis and Its Applications in Medical Imaging

Texture Analysis and Its Applications in Medical Imaging Edward J. Delp Purdue University School of Electrical and Computer Engineering Video and Image Processing Laboratory (VIPER) West Lafayette, Indiana, USA email: ace@ecn. purdue. edu http: //www. ece. purdue. edu/~ace Edward J. Delp Texture Analysis February 2000 Slide 1

What is Texture? • There is no good definition of texture in images! – IEEE: “texture is an attribute representing the spatial arrangement of the gray levels of the pixels in a region” – R. C. Gonzalez and P. Wintz (2 nd Edition): “we intuitively view this descriptor as providing a measure of properties such as smoothness, coarseness, and regularity” – W. K. Pratt: “several authors have attempted qualitatively to define texture” Edward J. Delp Texture Analysis February 2000 Slide 2

What is Texture? • J. K. Hawkins, “Textural Properties for Pattern Recognition, ” in Picture Processing and Pyschopictorics edited by B. S. Lipkins and A. Rosenfeld: – some “order” is repeated over a region which is large in comparison to the order’s size – the order consists of a in the nonrandom arrangement of elementary parts – the parts are roughly uniform entities having approximately the same dimensions everywhere in the region Edward J. Delp Texture Analysis February 2000 Slide 3

What is Texture? Edward J. Delp Texture Analysis February 2000 Slide 4

Textures • “Statistical” Textures • “Geometrical” Textures • Color Textures Edward J. Delp Texture Analysis February 2000 Slide 5

Image Texture • Texture Analysis – texture boundaries – texture properties • Texture Synthesis – generate synthetic texture • image compression • graphics Edward J. Delp Texture Analysis February 2000 Slide 6

Texture Measures - Moments • Model texture as a random process – basic concept - different textures have different statistical properties – variance – third (central) moment - skewness – fourth (central) moment - flatness • Problem - no spatial information used – Edward J. Delp Texture Analysis February 2000 Slide 7

Texture - Co-Occurrence Matrix • Estimate of the joint probability between two pixels in some neighborhood (2 d histogram) • A simple image (three gray levels): 00012 11011 22100 11020 00101 examine points in a region “one pixel to the right and one pixel below” Edward J. Delp Texture Analysis February 2000 Slide 8

Co-Occurrence Matrix • Let the image have m gray levels and form a m x m matrix – each entry in row i and column j is the number of pixels with gray level i below and with gray level j to the right: 421 232 020 Edward J. Delp Texture Analysis February 2000 Slide 9

Co-Occurrence Matrix • Divide each entry by the total number of point pairs • Measures used: – maximum value – entropy – total energy • Captures information about relative spatial position Edward J. Delp Texture Analysis February 2000 Slide 10

Texture - Other • Edge density • run length measures • frequency domain methods Edward J. Delp Texture Analysis February 2000 Slide 11

Multiresolution Decomposition • Transforms – Gaussian Pyramid – Morphological Pyramid – DCT – Wavelet Edward J. Delp Texture Analysis February 2000 Slide 12

Wavelet Transform Edward J. Delp gg 2 gh 2 hg 2 hh 2 gg 2 gh 2 Texture Analysis February 2000 Slide 13

Pyramid Representation Edward J. Delp Texture Analysis February 2000 Slide 14

Why Wavelets? • Tool for many multiresolution image processing and analysis techniques – rate scalable compression (JPEG 2000) – image watermarking – denoising – medical imaging – image and video databases – texture analysis Edward J. Delp Texture Analysis February 2000 Slide 15

Wavelet Texture Analysis • Excellent web site: http: //ua. ac. be/~visielab/wta. html Edward J. Delp Texture Analysis February 2000 Slide 16