Image Preprocessing Algorithm

One of the essential preprocessing operations implemented in any effective image processing algorithm is image segmentation or background elimination. In this study, all raw depth images were processed using the following procedure. First, image background removal was applied to the raw depth image data obtained using the image subtraction technique to remove the carcass background during image acquisition, as presented in Equation 1 (Li et al., 2002).

Where a(x,y) is the resultant image after it has been separated from its background, the original image is d(x,y), the background image is b(x−y).The threshold is T. Second, after establishing minimum and maximum thresholds, distance thresholding was performed based on depth image distance intensities to obtain the region of interest (ROI) according to Equation 2 (Jana, 2012; Okinda et al., 2018b).

Where b(x,y) is the resulting image following the removal of its background, Dmn is the minimum depth distance threshold, and Dmx is the maximum depth distance threshold. Third, Smoothening was performed on the b(x,y) using the Gaussian kernel filter (15 × 15 pixels zero mean) (Equation 3), then morphological opening by a disk structural element of size 9 pixels to remove small holes and obtain a clear depth image (Equation 4).

Where f(x,y)=f is the resultant filtered image, and hs,t=(12πσ2e−12(s2+t2σ2)) is the Gaussian filter kernel.

WhereI(x,y) is the resultant image after morphological opening, Mt is the structural element, (Mt)p is the translation of Mt by a point p.