Improved encode() error handling on wrong array shape

1. Improved encode() error handling on wrong array shape (silent crash on wrong array shape (#57)
2. Added an extra turbojpeg default path for Mac OS

Fixed lib paths missing on NetBSD

Fixed lib paths missing on NetBSD #55

fix incorrect logic on scale_with_quality()

scale_with_quality() returns empty data if decoding the image to be scaled triggers a warning. #53

Use libturbojpeg.so instead of libjpeg.so in FreeBSD.

Use libturbojpeg.so instead of libjpeg.so in FreeBSD. #51

v1.6.1

Map luminance level to dct coefficient #49

Here is an example to utilize @erikogabrielsson his contribution.

from PIL import Image
import numpy as np
import io
from turbojpeg import TurboJPEG

jpeg = TurboJPEG(r'C:\libjpeg-turbo64\bin\turbojpeg.dll')
in_file_path = r'in-1024x1024.jpeg'
in_file = open(in_file_path, 'rb')
in_data = in_file.read()
in_file.close()

# Crop the image to a larger size, fill extended area with white (1).
out_data = jpeg.crop_multiple(in_data, [(0, 0, 2048, 2048)], 1)[0]

# Write to file for analysis.
out_filepath = r'out.jpeg'
out_file = open(out_filepath, 'wb')
out_file.write(out_data)
out_file.close()

# Quick check that the end of the image array is RGB 255, 255, 255
np.asarray(Image.open(io.BytesIO(out_data)))

Support for multiple crop and fill background

crop_multiple() has been added to support for multiple crop and fill background #48

Add FreeBSD search path to avoid KeyError.

Add FreeBSD search path to avoid KeyError.  #47

Add default path to library in openSUSE systems

Add default path to library in openSUSE systems #45

Modify encode function to support 1-channel input

Modify encode function to support 1-channel input #42

Adding more code examples

Adding more code examples,

1. decoding input.jpg to BGR array with fast upsample and fast DCT. (i.e. fastest speed but lower accuracy)
2. using PyTurboJPEG with ExifRead to transpose an image if the image has an EXIF Orientation tag