A Composite Model of Wound Segmentation Based on Traditional Methods and Deep Neural Networks

Algorithm 1

Determine the dynamic thresholds.

Input: RGB channels of the image I

Output: CrA, CrB, the upper and lower Cr thresholds

of the skin with wound

RGB2YcbCr(I)

Cr unique(reshape((:,:, 3))

HCr Hist (Cr)

Sm [Cr(find(HCr = min(HCr(:, 1))), min(HCr(:,

1))]

Delete Sm(1,1:2), Sm(length(Sm(:,1)),1:2)

A [, ], B [, ]

for i,j1 to length(Cr)

do pp (sum(HCr(1:i))+

sum(HCr(length(Cr)-j: length(Cr)))

if pp > 5%

then lowi, highlength(Cr)-j, break

end

LCr (high) – Cr(low)

Lm0, pm0

while Lm >=15 & Lm <=30 & pm > Lm/L

do HA Sm(find(Sm(:,1)<=A &

Sm(:,1)>=A), 2)

HB Sm(find(Sm(:,1)<=B &

Sm(:,1)>=B), 2)

CrASm(find(Sm(:, 2)=min(HA)),1)

CrBSm(find(Sm(:, 2)=min(HB)),1)

LmCrB – CrA

pmsum(HCr(CrA:CrB))/sum(HCr)

AA+randi(,1,1), B B+randi(-

,,1,1)

end

return CrA, CrB