Skip to main content
Fig. 3 | Cancer Cell International

Fig. 3

From: Automatic model for cervical cancer screening based on convolutional neural network: a retrospective, multicohort, multicenter study

Fig. 3

Example image enhancement methods based on color and shape transformations. a Image rollover has alternating horizontal and vertical rollover components. Each time, one operation is randomly selected for image processing; b Processing of brightness image where the upper limit of the variation coefficient is set to 40, and an integer is randomly selected in the range [0–40] for brightness processing. c Processing of the saturation image where the range of the coefficient of change is set to [0.5,2], and the image color space is converted to the HSV space. A random value from the coefficient range is multiplied with the image in the saturation space each time. d Processing of the RGB color image where the coefficient of change is set to 60, the three R, G, B component values are randomly selected from [0,60]. In order to increase the image data variability, one or more of the four processing methods is are randomly selected each time

Back to article page