Edge Detection (with implementation on a GPU) And Text Recognition (if time permits) Jared Barnes Chris Jackson
Edge Detection ◦ Wikipedia: Identifying points in a digital image at which the image has discontinuities.
John Canny “A Computational Approach to Edge Detection”
1. Noise Removal 2. Image Gradient Computation 3. Non-Maximum Suppression 4. Hysteresis Thresholding
Gaussian Smoothing or Blurring A pixel is changed based on a weighted average of itself and its neighbors The number of neighbors (3x3, 5x5) and the relative weights can vary 3D Gaussian Distribution Normalized 2D Gaussian Approximation ages/articles/article-9/2d_distribution.gif s/spring08-22C251/homework/canny.pdf
Too much About right /tumblr_inline_mhcv1l0EZB1qz4rgp.png content/uploads/2013/03/gaussian-blur-thumbnail.jpg SpottySmooth
GxGx GyGy Sobel Operator (2 kernels) Then round to: 0° =←→ 90°=↑↓ 45°=↗↙ 135°=↘↖
X Gradient (Horizontal Edges) Y Gradient (Vertical Edges)
Make edges exactly one pixel thick Look at the gradient magnitude of your 2 neighbors in the direction of your angle Example 2 Angle = 0° ←→ Example 1 Angle = 135° ↘↖ Keep it! Kill it!
Thick Edges (Gradient Magnitude) Thin Edges (Gradient Magnitude)
Two thresholds are better than one! If a pixel’s value is above T high, it’s an edge. If a pixel’s value is below T low, it’s not an edge. If a pixel’s value is between T high and T low, it might be an edge (provided it is connected to an actual edge) T high = 45T low =
1. Smooth image to reduce noise 2. Calculate X & Y derivatives to get edges 3. Thin all edge widths to 1 pixel 4. Remove weak, unconnected edges (ta da!)
How do we parallelize the Canny Edge Detector?
Convolution – Independent of order ImageKernel Element-wise Multiplication 230 Sum All Values 11 Divide by Kernel Sum 11
Convolve a Gaussian Kernel with the image Each GPU core can convolve each pixel in the image individually with the Gaussian Kernel One thread per pixel, each performing 9 multiplies, 9 adds, and 1 division Embarrassingly Parallel with huge speedup
Convolve two Sobel Kernels with the image Wait, convolution again? Same as previous step – we can even reuse the convolution function!
Comparing 3 pixel gradient magnitudes and clearing the middle pixel or leaving it alone Similar to convolution… but simpler! Each GPU thread owns a pixel: 1.Check gradient angle of pixel 2.Compare this pixel’s magnitude with two neighbors in the direction of its angle 3.If I’m greater than those neighbors, leave me alone; otherwise, mark me as “not an edge” Less speedup than steps 1 and 2
Mark pixels > T high as strong edges Mark pixels < T low as not edges Mark remaining pixels as weak edges if they connect to a strong edge Typically implemented with recursion Each thread with a weak-edge pixel looks at nearest 2 neighbors to find a strong-edge pixel With identical algorithms on CPU and GPU, speedup is marginal (memory accesses, not much processing)
Wikipedia: The mechanical or electronic conversion of images of printed text into computer-readable text.
Label Connected Components Look For Letters Adjust for disconnected letters HELLO WORLD HELLOWORLDHELLOWORLD E F ? ü j i
Create a list of components in the image A component is simply a set of connected edges 1. Label each edge pixel with a unique component ID 2. Examine each pixel’s 8 touching neighbors and set that pixel’s ID to the smallest neighbor ID 3. Repeat step 2 until no pixel IDs are changed
Uhh… what’s a letter? How do we know it’s a letter? How does the computer know it’s a letter?
Letters are represented by a vector of numbers indicating the ratio of black pixels to white pixels in each division of the letter-image. A
Compute how closely each labelled component matches each letter in your alphabet The component is then marked with whichever letter it most closely matches
Letters like ‘i’ and ‘j’ have floating parts Sometimes edge detection may accidentally break up a letter A letter vector should then get an additional property indicating vertical discontinuity T E R V EL EO R………C T 0/1