Converting .mat file extension image to .jpg via python

Question

I'm currently trying to converting the images from a .mat file to .jpg file downloaded from this site- BrainTumorDataset. All the files contained in the directory are .mat files, now I want to convert all the files in .jpg format via python for making a project(Brain Tumor Classification using Deep Neural Net) via CNN. I searched in google but then I didn't get anything from there, only some topics on how to load .mat file in python but that also didn't help me. I found an answer in StackOverflow but this didn't work with this dataset and also the answer is for loading .mat image in python but I want to convert .mat images in .jpg format.

Answer 1

I managed to convert one image, use a loop to convert all.

Please read the comments.

import matplotlib.pyplot as plt
import numpy as np
import h5py
from PIL import Image

#reading v 7.3 mat file in python
#https://stackoverflow.com/questions/17316880/reading-v-7-3-mat-file-in-python

filepath = '1.mat';
f = h5py.File(filepath, 'r') #Open mat file for reading

#In MATLAB the data is arranged as follows:
#cjdata is a MATLAB struct
#cjdata.image is a matrix of type int16

#Before update: read only image data.   
####################################################################
#Read cjdata struct, get image member and convert numpy ndarray of type float
#image = np.array(f['cjdata'].get('image')).astype(np.float64) #In MATLAB: image = cjdata.image
#f.close()
####################################################################

#Update: Read all elements of cjdata struct
####################################################################
#Read cjdata struct
cjdata = f['cjdata'] #<HDF5 group "/cjdata" (5 members)>

# In MATLAB cjdata = 
# struct with fields:
#   label: 1
#   PID: '100360'
#   image: [512×512 int16]
#   tumorBorder: [38×1 double]
#   tumorMask: [512×512 logical]

#get image member and convert numpy ndarray of type float
image = np.array(cjdata.get('image')).astype(np.float64) #In MATLAB: image = cjdata.image

label = cjdata.get('label')[0,0] #Use [0,0] indexing in order to convert lable to scalar

PID = cjdata.get('PID') # <HDF5 dataset "PID": shape (6, 1), type "<u2">
PID = ''.join(chr(c) for c in PID) #Convert to string https://stackoverflow.com/questions/12036304/loading-hdf5-matlab-strings-into-python

tumorBorder = np.array(cjdata.get('tumorBorder'))[0] #Use [0] indexing - convert from 2D array to 1D array.

tumorMask = np.array(cjdata.get('tumorMask'))

f.close()
####################################################################

#Convert image to uint8 (before saving as jpeg - jpeg doesn't support int16 format).
#Use simple linear conversion: subtract minimum, and divide by range.
#Note: the conversion is not optimal - you should find a better way.
#Multiply by 255 to set values in uint8 range [0, 255], and covert to type uint8.
hi = np.max(image)
lo = np.min(image)
image = (((image - lo)/(hi-lo))*255).astype(np.uint8)

#Save as jpeg
#https://stackoverflow.com/questions/902761/saving-a-numpy-array-as-an-image
im = Image.fromarray(image)
im.save("1.jpg")

#Display image for testing
imgplot = plt.imshow(image)
plt.show()

---

Note:
Each mat file contains a struct named cjdata.
Fields of cjdata struct:

cjdata = 

struct with fields:

      label: 1
        PID: '100360'
      image: [512×512 int16]
tumorBorder: [38×1 double]
  tumorMask: [512×512 logical]

When converting images to jpeg, you are loosing information...

Answer 2

Here is how you can use a loop to convert all images.

from os import path
import os
from matplotlib import pyplot as plt
import numpy as np
import h5py
from PIL import Image
import re
import sys
from glob import glob


dir_path = path.dirname(path.abspath(__file__))
path_to_mat_files = path.join(dir_path, "*.mat")
found_files = glob(path_to_mat_files, recursive=True)
total_files = 0


def convert_to_png(file: str, number: int):
    global total_files
    if path.exists(file):
        print(file, "already exist\nSkipping...")
    else:
        h5_file = h5py.File(file, 'r')
        png = file[:-3] + "png"
        cjdata = h5_file['cjdata']
        image = np.array(cjdata.get('image')).astype(np.float64)
        label = cjdata.get('label')[0,0]
        PID = cjdata.get('PID')
        PID = ''.join(chr(c) for c in PID)
        tumorBorder = np.array(cjdata.get('tumorBorder'))[0]
        tumorMask = np.array(cjdata.get('tumorMask'))
        h5_file.close()
        hi = np.max(image)
        lo = np.min(image)
        image = (((image - lo)/(hi-lo))*255).astype(np.uint8)
        im = Image.fromarray(image)
        im.save(png)
        os.system(f"mv {png} {dir_path}\\png_images")#make sure folder png_images exist
        total_files += 1
        print("saving", png, "File No: ", number)
        
for file in found_files:
    if "cvind.mat" in file:
        continue
    convert_to_png(file, total_files)
print("Finished converting all files: ", total_files)