Optimized Periocular Template Selection for Human Recognition

Algorithm 1

Skin_Detection.

Require: 𝐼: RGB face image of size 𝑚×𝑛

Ensure: 𝑆: Binary face image indicating skin-map

 (1) Convert RGB image 𝐼 to 𝑌𝐶_𝑏𝐶_𝑟 color space

 (2) Normalize _𝐼𝑌_𝑖,𝑗 to [0,255] where_𝐼𝑌_𝑖,𝑗 denotes 𝑌 value for the pixel (𝑖,𝑗)

 (3) Compute the average luminance value of image 𝐼 as

  _𝐼𝑌_avg=1𝑚𝑛^𝑚𝑛∑_𝑖=1∑_𝐼𝑗=1𝑌_𝑖,𝑗

 (4) Brightness compensated image _𝐼𝐶^′ is obtained as _𝐼𝐶^′_𝑖,𝑗=_𝐼𝑅^′_𝑖,𝑗,_𝐼𝐺^′_𝑖,𝑗,_𝐼𝐵_𝑖,𝑗

  where, _𝐼𝑅^′_𝑖,𝑗=(_𝐼𝑅_𝑖,𝑗)^𝜏 and _𝐼𝐺^′_𝑖,𝑗=(_𝐼𝐺_𝑖,𝑗)^𝜏 and

  𝜏=1.5,if_𝐼𝑌_avg<640.7,if_𝐼𝑌_avg>1901,otherwise

 (5) The skin map 𝑆 is detected from _𝐼𝐶^′ as

  𝑆_𝑖,𝑗=𝑅0,if^𝑖,𝑗+1𝐺_𝑖,𝑗𝑅+1>1.08,^𝑖,𝑗+1𝐵_𝑖,𝑗+1>1.08,𝐺_𝑖,𝑗>30,𝐺_𝑖,𝑗<1401,otherwise

  where 𝑆_𝑖,𝑗=0 indicates skin region and 𝑆_𝑖,𝑗=1 indicates non-skin regions

 (6) return  𝑆