Create a normal texture from in Blender image.

make_normal.py

# ##### BEGIN GPL LICENSE BLOCK #####
#
#  This program is free software; you can redistribute it and/or
#  modify it under the terms of the GNU General Public License
#  as published by the Free Software Foundation; either version 2
#  of the License, or (at your option) any later version.
#
#  This program is distributed in the hope that it will be useful,
#  but WITHOUT ANY WARRANTY; without even the implied warranty of
#  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
#  GNU General Public License for more details.
#
#  You should have received a copy of the GNU General Public License
#  along with this program; if not, write to the Free Software Foundation,
#  Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
#
# ##### END GPL LICENSE BLOCK #####
import bpy
import numpy as np
import mathutils

img_name = 'bump.png'
strength = 10
sample = 24 # 8 or 24
useEase = True
exponent = 6 # [2, 6] higher value, higher contrast

# [min, max)
def clamp(x, min, max):
    if x < 0:
        x = 0
    elif x >= width:
        x = width-1
    return x
        
def get_pixel(img, width, height, x, y):
    x = clamp(x, 0, width)
    y = clamp(y, 0, height)
    
    idx = 4*(y*width + x)
    return img[idx], img[idx + 1], img[idx + 2]

def get_r(img, width, height, x, y):
    x = clamp(x, 0, width)
    y = clamp(y, 0, height)
    
    return img[4*(y*width + x)]
    
def set_pixel(img, width, height, x, y, r, g, b):
    idx = 4*(y*width + x)
    img[idx]=r
    img[idx+1] = g
    img[idx+2] = b
    img[idx+3] = 1.0
    
def calc_normal(l, r, t, b):
    du = (r-l)*strength
    dv = (b-t)*strength
    n = mathutils.Vector((dv, du, 1))
    n.normalize()
    return n

img_src = bpy.data.images[img_name]
img = np.array(img_src.pixels[:], dtype=np.float16)

# to grayscale
gray = 0.299 * img[0::4] + 0.587 * img[1::4] + 0.114 * img[2::4]

#easing
def ease(gray):
    if gray < 0.5:
        return 0.5*((2*gray)**exponent)
    else:
        return 1-0.5*((2-2*gray)**exponent)
        
if useEase:
    ease_func = np.frompyfunc(ease, 1, 1)
    gray = ease_func(gray)

img[0::4] = gray

#calc
width, height = img_src.size
dst = np.empty(4*width*height, dtype=np.float16)

if sample == 8:
    #8 sample
    for y in range(height):
        for x in range(width):
            l = get_r(img, width, height, x-1, y)
            r = get_r(img, width, height, x+1, y)
            t = get_r(img, width, height, x, y-1)
            b = get_r(img, width, height, x, y+1)        
            n1 = calc_normal(l, r, t, b)

            l = get_r(img, width, height, x-1, y-1)
            r = get_r(img, width, height, x+1, y+1)
            t = get_r(img, width, height, x+1, y-1)
            b = get_r(img, width, height, x-1, y+1)
            n2 = calc_normal(l, r, t, b)
            
            n = 0.25*(n1 + n2)
            n = n + mathutils.Vector((0.5, 0.5, 0.5))
            set_pixel(dst, width, height, x, y, n.x, n.y, n.z)
            
elif sample == 24:
    # 24 sample
    for y in range(height):
        for x in range(width):
            l = get_r(img, width, height, x-1, y)
            r = get_r(img, width, height, x+1, y)
            t = get_r(img, width, height, x, y-1)
            b = get_r(img, width, height, x, y+1)        
            n1 = calc_normal(l, r, t, b)
            
            l = get_r(img, width, height, x-1, y-1)
            r = get_r(img, width, height, x+1, y+1)
            t = get_r(img, width, height, x+1, y-1)
            b = get_r(img, width, height, x-1, y+1)
            n2 = calc_normal(l, r, t, b)
            
            l = get_r(img, width, height, x-2, y-2)
            r = get_r(img, width, height, x+2, y+2)
            t = get_r(img, width, height, x+2, y-2)
            b = get_r(img, width, height, x-2, y+2)        
            n3 = calc_normal(l, r, t, b)
            
            l = get_r(img, width, height, x-2, y)
            r = get_r(img, width, height, x+2, y)
            t = get_r(img, width, height, x, y-2)
            b = get_r(img, width, height, x, y+2)        
            n4 = calc_normal(l, r, t, b)
            
            l = get_r(img, width, height, x-2, y-1)
            r = get_r(img, width, height, x+2, y+1)
            t = get_r(img, width, height, x+2, y-1)
            b = get_r(img, width, height, x-2, y+1)        
            n5 = calc_normal(l, r, t, b)
            
            l = get_r(img, width, height, x-1, y-2)
            r = get_r(img, width, height, x+1, y+2)
            t = get_r(img, width, height, x+1, y-2)
            b = get_r(img, width, height, x-1, y+2)        
            n6 = calc_normal(l, r, t, b)
            
            n = 0.08333333*(n1 + n2 + n3 + n4 + n5 + n6)
            n = n + mathutils.Vector((0.5, 0.5, 0.5))
            set_pixel(dst, width, height, x, y, n.x, n.y, n.z)

#to image
new = bpy.data.images.new('normal_'+img_name, width, height, alpha=True)
new.pixels = dst