Image Processing Rule 1: Always save your primary data Rule 2: Be able to describe quantitatively what you have done 3 types of operations вЂў Point operations: operate on a pixel-by-pixel basis вЂў Neighborhood operations: operate on a small group of adjacent pizels вЂў Reciprocal space operations: deal with imagewide patterns and chartacteristics Point operations (Black-and-White) http://homepages.inf.ed.ac.uk/rbf/HIPR2/pnto ps.htm Thresholding Gamma Curves Histogram Equalization Point Operations (Color) вЂў Color maps вЂ“ RGB (More common in scientific imaging) вЂ“ CMYK (Printers, etc) вЂў Color Balance вЂў HSV Shading correction вЂў Two sources of shading variation in images вЂ“ Dye binding to background: Image subtraction is appropriate вЂ“ Camera/light source nonuniformity: Image division is appropriate вЂў Image subtraction вЂў Image division: Divide data image by blank image and normalize вЂў If you donвЂ™t have a blank image: erode features and smooth to derive background (вЂњflatten imageвЂќ in Image Pro Plus) Geometric correction вЂў Geometric distortion: pincushion and barrel distortion вЂў Geometric distortion: trapezoidal distortion вЂў Tiepoints: set of points with known geometric relationships to each other вЂў Set up a matrix of actual geometric positions in the image as a function of pixel coordinates вЂў Interpolate beween nonintegrap pixels positions to get square pixels. Neigborhood operations вЂ“ allow feature extraction вЂў Convolution operator: a matrix that applies a kernel (say 3X3) to every point in the image вЂў Replaces the central point by the resultant of multiplying that 3X3 matrix by neighboring pixels Molecular Expressions Web Site Averaging kernel вЂў Replace central point with average of neighborhood 1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9 Most software packages do the normalization automatically, so you can use вЂњ1вЂќвЂ™s instead of вЂњ1/9вЂќвЂ™s Smoothing kernel вЂў Gaussian (3X3) 1 4 1 4 12 4 1 4 1 вЂў (5x5) 1 2 3 2 1 2 7 11 7 2 3 11 17 11 3 2 7 11 7 2 1 2 3 2 1 Sharpening kernels вЂў Laplacian -1 -1 -1 -1 8 -1 -1 -1 -1 Approximates a Laplacian operator, which replaces the central value with the differential in x and y Directional kernels вЂў Vertical edge -1 0 1 -1 0 1 -1 0 1 Average in vertical direction Difference in horizontal direction вЂў Diagonal edge 2 1 0 1 0 -1 0 -1 -2 Complex neighborhood operations вЂў Median filter: replace central pixel with median of neighborhood вЂ“ Very effective at removing вЂњshot noiseвЂќ вЂў Roberts cross: 2 perpendicular directional filter вЂў Sobel: вЂ“ Calculate derivatives in 2 perpendicular directions вЂ“ Replace central magnitude with в€љ ((ОґB/ Оґx)2 + (ОґB/ Оґy)2 ) вЂў Kirsch: Apply each of 8 directional filters, and replace central value with maximum Complex neighborhood operations вЂў Olympic filter: in each 5X5 neighborhood. Ignote the brightest and darkest 4. Replace the central value with the average of the remaining 17 вЂў Top hat: replace values greater than the average of a neighborhood by the average for that neigborhood вЂў Gray scale opening вЂ“ First pass: replace central pixel with brightest neighbor вЂ“ Second pass: replace pixel with darkest neigbor вЂ“ Net effect: dilation of dark features, and erosion of bright Hybrid вЂ“ sharpening by difference of Gaussians (unsharp mask) вЂў Apply 2 different size Gaussians to same image вЂў Subtract smaller from larger Gaussian filtered result вЂў Unsharp masking вЂ“ Photographically: вЂў Image in and out of focus вЂў Invert out-of-focus вЂў Mat reversed image with in-focus вЂ“ Digitally: subtract blurred from unblurred вЂ“ http://micro.magnet.fsu.edu/primer/java/digitalimagi ng/processing/unsharpmask/index.html Character recognition вЂў Instead of regular convolution masks, use masks that represent characters in the image вЂў You get a вЂњhitвЂќ, or high match, whenever the mask matches the character! вЂў However, the characters must ba aligned, undistorted, etc. Automatic number plate recognition Wikipedia: Automatic number plate recognition Algorithms for ANPR вЂў вЂў вЂў вЂў вЂў вЂў вЂў вЂў There are six primary algorithms that the software requires for identifying a licence plate: Plate localisation вЂ“ responsible for finding and isolating the plate on the picture Plate orientation and sizing вЂ“ compensates for the skew of the plate and adjusts the dimensions to the required size Normalisation вЂ“ adjusts the brightness and contrast of the image Character segmentation вЂ“ finds the individual characters on the plates Optical character recognition Syntactical/Geometrical analysis вЂ“ check characters and positions against country specific rules The complexity of each of these subsections of the program determines the accuracy of the system. During the third phase (normalisation) some systems use edge detection techniques to increase the picture difference between the letters and the plate backing. A median filter may also be used to reduce the visual "noise" on the image. -ibid General object recognition вЂў How do we recognize specific objects (such as tanks in aerial images) using machine vision вЂ“ Problem of orientation: any orientation may present itself вЂў First, scan image with circularly averaged structures вЂў Then, scan again with specific orietations вЂў Highly computationally expensive, and not terribly effective вЂў We can often do better with Fourier transform techniques. Fourier Transform Image Processing вЂў Any periodic object can be represented by a summation of a series of cosine waves вЂў The Operation of Fourier transformation of an image replaces the image (real space) be a series of amplitudes and frequencies of the cosine waves that make it up вЂў Fourier space is also referred to as frequency space вЂў If there are repeats in the stucture at specific frequencies, these will appear as peaks in Fourier space Fourier Transform Image Processing вЂў High- and low-pass filters вЂў By enhancing or supressing specific frequencies, we can enhance or suppress periodic structures within the image Molecular Expressions Java simulation Examples вЂў Nuclear pore complex вЂ“ Markham rotation вЂ“ Fourier transform вЂў Removal of halftone screen noise Dangers of Fourier transforms вЂў Can introduce periodicities where none are present вЂў Edge effects

1/--страниц