A Data-Driven and Biologically Inspired Preprocessing Scheme to Improve Visual Object Recognition

Algorithm 1

The steps of Image preprocessing.

	Input: raw image (I)
	Output: saliency region (S)
	Step 1: RGB to CIE LAB color space conversion:(1)  RGB space is transformed to XYZ space according to formula (1)(2)  XYZ space is transformed to LAB space according to formula (2)
	Step 2: RGC response’s simulation:(1)  The L, A, and B channels are convolved with the Gaussian function based on the midget cells’ in the human retina (the standard deviation of this Gaussian function is considered according to Table 1)
	Step 3: binary map generation(1)  By selecting an appropriate threshold, the Responses are converted to binary images (the threshold is chosen based on the mean value of each gray image)
	Step 4: the new image representation(1)  The mask operation is done by using the Binary maps in the previous stage
	Step 5: ROI extraction(1)  All images are combined by using entropy quantity according to formula (9) The coefficients are calculated according to formula (10)