AF-SENet: Classification of Cancer in Cervical Tissue Pathological Images Based on Fusing Deep Convolution Features

. 2020 Dec 27;21(1):122. doi: 10.3390/s21010122

Algorithm 1 Image Enhancement Processing

Input:

Y_{I}

Output:

Y_{o}

Y_{I}

is the original image matrix,

Y_{o}

is the enhanced image matrix.

2 FOR p = 1: s //s is the number of image samples

3 Implement random cropping based on grayscale matching for

y_{s}^{n}

4 (i) Perform random cropping, according to Formulas (4) and (5):

y_{p}^{m} = C r o p (y_{s}^{n}, s i z e)

6 (ii) Determine whether the following conditions are met:

D (i, j) < D_{T}

Y_{o} = Y_{o} \cup y_{p}^{m}

8 Randomly shift the randomly cropped image tensor

y_{p}^{m}

according to Equation (6):

y_{p}^{k + 1} = W a r p (y_{p}^{m}) t h e n Y_{o} = Y_{o} \cup y_{p}^{k + 1}

10 Randomly rotate

y_{p}^{m}

according to Formula (7):

y_{p}^{k + 2} = R o t a t i o n (y_{p}^{m}) t h e n Y_{o} = Y_{o} \cup y_{p}^{k + 2}

12 Randomly scale

y_{p}^{m}

according to Formula (8):

y_{p}^{k + 3} = Z o o m (y_{p}^{m}) t h e n Y_{o} = Y_{o} \cup y_{p}^{k + 3}

14 Randomly adjust the brightness of

y_{p}^{m}

according to Equation (9):

y_{p}^{k + 4} = B r i g h t (y_{p}^{m}) t h e n Y_{o} = Y_{o} \cup y_{p}^{k + 4}

16 Normalize the enhanced, small-sized image according to Equation (10):

Y_{o} = Y_{o} / 255.0

18 END

19 Return

Y_{o}