Edge Detection: A Collection of Pixel based Approach for Colored Images

摘要

The existing traditional edge detection algorithms process a single pixel onan image at a time, thereby calculating a value which shows the edge magnitudeof the pixel and the edge orientation. Most of these existing algorithmsconvert the coloured images into gray scale before detection of edges. However,this process leads to inaccurate precision of recognized edges, thus producingfalse and broken edges in the image. This paper presents a profile modellingscheme for collection of pixels based on the step and ramp edges, with a viewto reducing the false and broken edges present in the image. The collection ofpixel scheme generated is used with the Vector Order Statistics to reduce theimprecision of recognized edges when converting from coloured to gray scaleimages. The Pratt Figure of Merit (PFOM) is used as a quantitative comparisonbetween the existing traditional edge detection algorithm and the developedalgorithm as a means of validation. The PFOM value obtained for the developedalgorithm is 0.8480, which showed an improvement over the existing traditionaledge detection algorithms.