A Novel Approach to Line Detection using Image Integration Method
Daniel Wei Lin, Bo Cun Sun · 2015
We developed a novel line detection method that requires no filters to remove noise on images. The figure above is a sample of the test image we used to test our program on. It is a 100 x 100 pixel image with ten lines. The lines are generated using Bresenham Line Generation Algorithm. The figure above is the correlation map produced from the test image. The horizontal axis represent the length of the line and the vertical axis represents the angle of the line. We selected the angle between 0 to 180 degrees and the length from –50 to 50 pixels for our test cases. After generating the correlation map, we filtered it to remove peaks below 40% of the highest value. The orange lines represent the lines that were on our test image. The green lines represent the lines detected by our algorithm. We decided to make the algorithm plot the detected lines on the original image without noise to demonstrate the accuracy of our method.