I am trying to make stamped metal effect into an image, so I go with applying bevel effect like in photoshop and I came across this question. This image for reference about what am I trying to achieve.

import cv2
import numpy as np
from matplotlib import pyplot as plt
from matplotlib import cm
from PIL import Image
import scipy

def nothing(x):
    pass

if __name__ == '__main__':
    mask = np.zeros((200,400,1), np.uint8)
    mask = cv2.rectangle(mask, (30,30), (180,180), 255, -1)
    mask = cv2.circle(mask, (30,30), 50, 0, -1)
    mask = cv2.circle(mask, (180,180), 50, 0, -1)
    mask = cv2.circle(mask, (300,100), 75, 255, -1)
    mask = cv2.circle(mask, (300,100), 25, 0, -1)
    mask = cv2.rectangle(mask, (313,88), (338,110), 0, -1)

    cv2.namedWindow('Bevel')

    while(True):
        light_angle = 0
        depth = 7
        alfa = 1.0
        shift = 128

        kernel_dist = np.ones((3,3), np.uint8)
        dist = cv2.morphologyEx(mask, cv2.MORPH_GRADIENT, kernel_dist)
        #dist = cv2.Canny(mask, 50,150)
        dist = 255-dist
        dist = cv2.distanceTransform(dist, cv2.DIST_L2, cv2.DIST_MASK_5)
        _, dist = cv2.threshold(dist, depth, depth, cv2.THRESH_TRUNC)
        dist = cv2.blur(dist, (3,3))
        #bevel= np.array(dist, dtype=np.uint8) # not used, only to show stuff
        #bevel= cv2.equalizeHist(bevel)
        d = np.array([[-1,-2,-3],
                    [-2, 0, 0],
                    [-3, 0, 1]], dtype=np.float32)

        d = scipy.ndimage.rotate(d,light_angle)
        kernel = d - np.mean(d)
        dist2 = cv2.filter2D(dist, cv2.CV_32F, kernel)
        maxval = cv2.minMaxLoc(dist2)
        dist2 = 128*dist2/maxval[1]

        bevel2 = dist2
        bevel2 = shift - bevel2 * alfa 
        bevel2 = np.multiply(bevel2.reshape(mask.shape), mask)/255
        
        bg = np.zeros((200,400,3), np.float32)
        bg = cv2.rectangle(bg, (30,30), (180,180), (150,150,0), -1)
        bg = cv2.circle(bg, (30,30), 50, (0,0,0), -1)
        bg = cv2.circle(bg, (180,180), 50, (0,0,0), -1)
        bg = cv2.circle(bg, (300,100), 75, (0,200,0), -1)
        bg = cv2.circle(bg, (300,100), 25, (0,0,0), -1)
        bg = cv2.rectangle(bg, (313,88), (338,110), (0,0,0), -1)

        bg_hsv = cv2.cvtColor(bg, cv2.COLOR_BGR2HSV)
        h,s,v = cv2.split(bg_hsv)
        v_blend = cv2.add(v,bevel2) #here merging bevel map and image via V channel
        bg_hsv_new = cv2.merge([h,s,v_blend])
        cv2.imshow('Bevel', cv2.cvtColor(bg_hsv_new, cv2.COLOR_HSV2BGR).astype(np.uint8))
        
        if cv2.waitKey(10) & 0xFF == ord('q'):
            break

The question is :

1. Is matrix/method that used on referenced question has an actual name? I couldn’t find any source that explain how said matrix/method work and how to tune it (such as strength of light, approach angle etc.)

2. Is my implementation of merging bevel map and actual image correct? The result become cool looking reflective metal effect, but that achieved after reducing parameter shift into 0 (where on referenced question used constant 128). If not, how to correctly use bevel map that produced?

Result :

Shift = 128

Shift = 0