Welcome, guest | Sign In | My Account | Store | Cart

Saving a numpy array to png image using fitz / MuPDF (PyMuPDF) (Python recipe) by Harald Lieder
ActiveState Code (http://code.activestate.com/recipes/580630/)

If a pixel array is defined in numpy, it can be saved as a PNG image VERY fast if you use PyMuPDF.

Python, 30 lines
Download
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
#! /usr/bin/python
# -*- coding: utf-8 -*-
import numpy as np
import fitz             # this PyMuPDF

#==============================================================================
# Create any height*width*4 RGBA pixel area using numpy and then use fitz
# to save it as a PNG image.
# This is usually more than 10 times faster than pure python solutions
# like pypng and more than 2 times faster than PIL / Pillow.
#==============================================================================
height = 108            # choose whatever
width  = 192            # you want here 
image  = np.ndarray((height, width, 4), dtype=np.uint8) # data type uint8 is required!

#==============================================================================
# Fill the array. Each 4-byte-tuple represents a pixel with (R, B, G, A).
# We have chosen A = 255 here (non transparent).
#==============================================================================
for i in range(height):
    for j in range(width):
        # just demoing - choose color components as you like them
        image[i, j] = np.array([i%256, j%256, 200, 255], dtype=np.uint8)

#==============================================================================
# create a string from the array and output the picture
#==============================================================================
samples = image.tostring()
pix = fitz.Pixmap(fitz.csRGB, width, height, samples)
pix.writePNG("test.png")
Tags: fitz, image, mupdf, numpy, png, pymupdf