Need to crop thousands of images / sprites while preserving midpoint - image

Question

I have thousands of character sprites (though, really they're just PNG files with transparency) which were saved out from Maya.
These are character animations, with almost all of them being 12 frames each. Each animation is rendered from each of the 8 cardinal directions.
In each animation, the character is already centered in the frame (this is how it's exported from Maya); however, each image has a bunch of white/transparent space around it.
I need to batch crop these sprites horizontally from the outside -> in order to preserve the character's midpoint in the frame. The reason for this is if it's not preserved, as the game swaps from one animation to the next, the character would appear to move or jump around. So, it's important the character is horizontally centered (using the midpoint of their body) in the frame across all their animations.
Additionally... ideally the character is always at or very near the bottom edge of the sprite.
For the coup de grace... the final cropped image needs to be an even number, or divisible by 2, for game engine reasons.
Is there any way to at least partially, or entirely, automate this? Or are there any programs that someone could recommend to help with the automation of this; free or paid.

Answer 1

I dunno yet if a Python-based answer suits your but here is what I got, with a bit of work your can copy and paste this script to your script editor in Maya or execute it in your Python interpreter.
This script requires the PIL library to be installed.
To install it on Maya, you can follow this link Mistermatti's blog. Unzip the file and paste the PIL folder in your maya install here: Maya20XX\Python\Lib\site-packages
This works on Maya2014, havn't tested on other versions tho.
from PIL import Image
import time
start_time = time.time()
#parses the image to find the first pixel on the x axis which is not blank
def parseXImage(pixelMap, resWidth, resHeight):
for x in range(0, int(resWidth/2)):
for y in range(0, int(resHeight/2)):
if (pixelMap[x,y] != (0, 0, 0, 0)) or \
(pixelMap[x,resHeight-y-1] != (0, 0, 0, 0)) or \
(pixelMap[resWidth-x-1,y] != (0, 0, 0, 0)) or \
(pixelMap[resWidth-x-1,resHeight-y-1] != (0, 0, 0, 0)):
return x
return None
#parses the image to find the first pixel on the y axis which is not blank
def parseYImage(pixelMap, resWidth, resHeight):
topFound = False
bottomFound = False
yTop = None
yBottom = None
for y in range(0, int(resHeight/2)):
for x in range(0, int(resWidth/2)):
if not topFound:
if (pixelMap[x,y] != (0, 0, 0, 0)) or \
(pixelMap[resWidth-x-1,y] != (0, 0, 0, 0)):
yTop = y
topFound = True
if not bottomFound:
if (pixelMap[x,resHeight-y-1] != (0, 0, 0, 0)) or \
(pixelMap[resWidth-x-1,resHeight-y-1] != (0, 0, 0, 0)):
yBottom = y
bottomFound = True
return yTop, yBottom
imageList = [r"Path\To\Image\Mod_Turn.001.png",
r"Path\To\Image\Mod_Turn.002.png",
r"Path\To\Image\Mod_Turn.003.png",
r"Path\To\Image\Mod_Turn.004.png",
r"Path\To\Image\Mod_Turn.005.png"]
#get images resolution
imagePil = Image.open(imageList[0])
resWidth, resHeight = imagePil.size #get the resolution of the first image in the list
imagePil.close()
valueWidthList = []
valueHeightTopList = []
valueHeightBottomList = []
for imageListIndex, imagePath in enumerate(imageList):
imagePil = Image.open(imagePath)
pixelMap = imagePil.load()
xValue = parseXImage(pixelMap, resWidth, resHeight) #now parse the image to find x
yTopValue, yBottomValue = parseYImage(pixelMap, resWidth, resHeight) #now parse the image to find y
valueWidthList.append(xValue) #Store it
valueHeightTopList.append(yTopValue) #Store it
valueHeightBottomList.append(yBottomValue) #Store it
imagePil.close()
xValueToCrop = min(valueWidthList) #Get the minimal value of X crop among all the entries
yTopValueToCrop = min(valueHeightTopList) #Get the minimal value of Y crop among all the entries
yBottomValueToCrop = min(valueHeightBottomList) #Get the minimal value of Y crop among all the entries
#This is the crop square we will use
cropSquare = (xValueToCrop, yTopValueToCrop, resWidth-xValueToCrop, resHeight-yBottomValueToCrop )
for imagePath in imageList:
imagePil = Image.open(imagePath)
#Replace Turn by something in the name of your image
imagePil.crop(cropSquare).save(imagePath.replace("Turn", "Turn_Cropped")) #Crop dans save the image
imagePil.close()
print("--- %s seconds ---" % (time.time() - start_time))
You need to change the imageList dict with your own paths to images and "Turn", "Turn_Cropped" at the very end of the script.
FYI, processing 300 images #720p took about 250 seconds. This can be improved.

Answer 2

Perhaps give TexturePacker a try https://www.codeandweb.com/texturepacker
I've used it before when dealing with sprites and find it quite nice. I believe it also has command-line functionality, though I haven't used it myself. And it's cross-platform.