I have an image and I want to create an edge histogram. I divide the image into 1100 image-blocks and try to find edge and its direction (horisontal, vertical, 45° diagonal,
I found the answer in this paper: Efficient Use of MPEG-7 Edge Histogram Descriptor by Won.
My goal was to find following edges:
Won divide each Image Block into 4 parts, calculate the average gray level in each of them and use the following coefficients:
We use this coefficients as follows and get 5 values:
Using thresholding we estimate each type of the edge:
program SetEdgeType(max, m_nd, m_h, m_v, m_d_45, m_d_135)
{
if (max < TEdge) then EdgeHisto(0)++
else
{
if (m_nd > T0) then EdgeHisto(1)++
if (m_h > T1) then EdgeHisto(2)++
if (m_v > T1) then EdgeHisto(3)++
if (m_d_45 > T2) then EdgeHisto(4)++
if (m_d_135 > T2) then EdgeHisto(5)++
}
endif
return(EdgeHisto)
}
Threshold values were selected by Savvas A. Chatzichristofis to be: TEdge=14, T0=0.68, T1=T2=0.98.