ImageTools is a class to deal with conversions from & to ROS Image, ROS CompressedImage and numpy.ndarray (cv2 image).

image_tools.py

#!/usr/bin/env python
import numpy as np
import cv2
from cv_bridge import CvBridge, CvBridgeError

"""
ROS ImageTools, a class that contains methods to transform
from & to ROS Image, ROS CompressedImage & numpy.ndarray (cv2 image).
Also deals with Depth images, with a tricky catch, as they are compressed in
PNG, and we are here hardcoding to compression level 3 and the default
quantizations of the plugin. (What we use in our robots).

Meanwhile image_transport has no Python interface, this is the best we can do.

Author: Sammy Pfeiffer <Sammy.Pfeiffer at student.uts.edu.au>
"""


class ImageTools(object):
    def __init__(self):
        self._cv_bridge = CvBridge()

    def convert_ros_msg_to_cv2(self, ros_data, image_encoding='bgr8'):
        """
        Convert from a ROS Image message to a cv2 image.
        """
        try:
            return self._cv_bridge.imgmsg_to_cv2(ros_data, image_encoding)
        except CvBridgeError as e:
            if "[16UC1] is not a color format" in str(e):
                raise CvBridgeError(
                    "You may be trying to use a Image method " +
                    "(Subscriber, Publisher, conversion) on a depth image" +
                    " message. Original exception: " + str(e))
            raise e

    def convert_ros_compressed_to_cv2(self, compressed_msg):
        np_arr = np.fromstring(compressed_msg.data, np.uint8)
        return cv2.imdecode(np_arr, cv2.IMREAD_COLOR)

    def convert_ros_compressed_msg_to_ros_msg(self, compressed_msg,
                                              encoding='bgr8'):
        cv2_img = self.convert_ros_compressed_to_cv2(compressed_msg)
        ros_img = self._cv_bridge.cv2_to_imgmsg(cv2_img, encoding=encoding)
        ros_img.header = compressed_msg.header
        return ros_img

    def convert_cv2_to_ros_msg(self, cv2_data, image_encoding='bgr8'):
        """
        Convert from a cv2 image to a ROS Image message.
        """
        return self._cv_bridge.cv2_to_imgmsg(cv2_data, image_encoding)

    def convert_cv2_to_ros_compressed_msg(self, cv2_data,
                                          compressed_format='jpg'):
        """
        Convert from cv2 image to ROS CompressedImage.
        """
        return self._cv_bridge.cv2_to_compressed_imgmsg(cv2_data,
                                                        dst_format=compressed_format)

    def convert_ros_msg_to_ros_compressed_msg(self, image,
                                              image_encoding='bgr8',
                                              compressed_format="jpg"):
        """
        Convert from ROS Image message to ROS CompressedImage.
        """
        cv2_img = self.convert_ros_msg_to_cv2(image, image_encoding)
        cimg_msg = self._cv_bridge.cv2_to_compressed_imgmsg(cv2_img,
                                                            dst_format=compressed_format)
        cimg_msg.header = image.header
        return cimg_msg

    def convert_to_cv2(self, image):
        """
        Convert any kind of image to cv2.
        """
        cv2_img = None
        if type(image) == np.ndarray:
            cv2_img = image
        elif image._type == 'sensor_msgs/Image':
            cv2_img = self.convert_ros_msg_to_cv2(image)
        elif image._type == 'sensor_msgs/CompressedImage':
            cv2_img = self.convert_ros_compressed_to_cv2(image)
        else:
            raise TypeError("Cannot convert type: " + str(type(image)))
        return cv2_img

    def convert_to_ros_msg(self, image):
        """
        Convert any kind of image to ROS Image.
        """
        ros_msg = None
        if type(image) == np.ndarray:
            ros_msg = self.convert_cv2_to_ros_msg(image)
        elif image._type == 'sensor_msgs/Image':
            ros_msg = image
        elif image._type == 'sensor_msgs/CompressedImage':
            ros_msg = self.convert_ros_compressed_msg_to_ros_msg(image)
        else:
            raise TypeError("Cannot convert type: " + str(type(image)))
        return ros_msg

    def convert_to_ros_compressed_msg(self, image, compressed_format='jpg'):
        """
        Convert any kind of image to ROS Compressed Image.
        """
        ros_cmp = None
        if type(image) == np.ndarray:
            ros_cmp = self.convert_cv2_to_ros_compressed_msg(
                image, compressed_format=compressed_format)
        elif image._type == 'sensor_msgs/Image':
            ros_cmp = self.convert_ros_msg_to_ros_compressed_msg(
                image, compressed_format=compressed_format)
        elif image._type == 'sensor_msgs/CompressedImage':
            ros_cmp = image
        else:
            raise TypeError("Cannot convert type: " + str(type(image)))
        return ros_cmp

    def convert_depth_to_ros_msg(self, image):
        ros_msg = None
        if type(image) == np.ndarray:
            ros_msg = self.convert_cv2_to_ros_msg(image,
                                                  image_encoding='mono16')
        elif image._type == 'sensor_msgs/Image':
            image.encoding = '16UC1'
            ros_msg = image
        elif image._type == 'sensor_msgs/CompressedImage':
            ros_msg = self.convert_compressedDepth_to_image_msg(image)
        else:
            raise TypeError("Cannot convert type: " + str(type(image)))
        return ros_msg

    def convert_depth_to_ros_compressed_msg(self, image):
        ros_cmp = None
        if type(image) == np.ndarray:
            ros_cmp = self.convert_cv2_to_ros_compressed_msg(image,
                                                             compressed_format='png')
            ros_cmp.format = '16UC1; compressedDepth'
            # This is a header ROS depth CompressedImage have, necessary
            # for viewer tools to see the image
            # extracted from a real image from a robot
            # The code that does it in C++ is this:
            # https://github.com/ros-perception/image_transport_plugins/blob/indigo-devel/compressed_depth_image_transport/src/codec.cpp
            ros_cmp.data = "\x00\x00\x00\x00\x88\x9c\x5c\xaa\x00\x40\x4b\xb7" + ros_cmp.data
        elif image._type == 'sensor_msgs/Image':
            image.encoding = "mono16"
            ros_cmp = self.convert_ros_msg_to_ros_compressed_msg(
                image,
                image_encoding='mono16',
                compressed_format='png')
            ros_cmp.format = '16UC1; compressedDepth'
            ros_cmp.data = "\x00\x00\x00\x00\x88\x9c\x5c\xaa\x00\x40\x4b\xb7" + ros_cmp.data
        elif image._type == 'sensor_msgs/CompressedImage':
            ros_cmp = image
        else:
            raise TypeError("Cannot convert type: " + str(type(image)))
        return ros_cmp

    def convert_depth_to_cv2(self, image):
        cv2_img = None
        if type(image) == np.ndarray:
            cv2_img = image
        elif image._type == 'sensor_msgs/Image':
            image.encoding = 'mono16'
            cv2_img = self.convert_ros_msg_to_cv2(image,
                                                  image_encoding='mono16')
        elif image._type == 'sensor_msgs/CompressedImage':
            cv2_img = self.convert_compressedDepth_to_cv2(image)
        else:
            raise TypeError("Cannot convert type: " + str(type(image)))
        return cv2_img

    def convert_compressedDepth_to_image_msg(self, compressed_image):
        """
        Convert a compressedDepth topic image into a ROS Image message.
        compressed_image must be from a topic /bla/compressedDepth
        as it's encoded in PNG
        Code from: https://answers.ros.org/question/249775/display-compresseddepth-image-python-cv2/
        """
        depth_img_raw = self.convert_compressedDepth_to_cv2(compressed_image)
        img_msg = self._cv_bridge.cv2_to_imgmsg(depth_img_raw, "mono16")
        img_msg.header = compressed_image.header
        img_msg.encoding = "16UC1"
        return img_msg

    def convert_compressedDepth_to_cv2(self, compressed_depth):
        """
        Convert a compressedDepth topic image into a cv2 image.
        compressed_depth must be from a topic /bla/compressedDepth
        as it's encoded in PNG
        Code from: https://answers.ros.org/question/249775/display-compresseddepth-image-python-cv2/
        """
        depth_fmt, compr_type = compressed_depth.format.split(';')
        # remove white space
        depth_fmt = depth_fmt.strip()
        compr_type = compr_type.strip()
        if compr_type != "compressedDepth":
            raise Exception("Compression type is not 'compressedDepth'."
                            "You probably subscribed to the wrong topic.")

        # remove header from raw data, if necessary
        if 'PNG' in compressed_depth.data[:12]:
            # If we compressed it with opencv, there is nothing to strip
            depth_header_size = 0
        else:
            # If it comes from a robot/sensor, it has 12 useless bytes apparently
            depth_header_size = 12
        raw_data = compressed_depth.data[depth_header_size:]

        depth_img_raw = cv2.imdecode(np.frombuffer(raw_data, np.uint8),
                                     # the cv2.CV_LOAD_IMAGE_UNCHANGED has been removed
                                     -1)  # cv2.CV_LOAD_IMAGE_UNCHANGED)
        if depth_img_raw is None:
            # probably wrong header size
            raise Exception("Could not decode compressed depth image."
                            "You may need to change 'depth_header_size'!")
        return depth_img_raw

    def display_image(self, image):
        """
        Use cv2 to show an image.
        """
        cv2_img = self.convert_to_cv2(image)
        window_name = 'show_image press q to exit'
        cv2.imshow(window_name, cv2_img)
        # TODO: Figure out how to check if the window
        # was closed... when a user does it, the program is stuck
        key = cv2.waitKey(0)
        if chr(key) == 'q':
            cv2.destroyWindow(window_name)

    def save_image(self, image, filename):
        """
        Given an image in numpy array or ROS format
        save it using cv2 to the filename. The extension
        declares the type of image (e.g. .jpg .png).
        """
        cv2_img = self.convert_to_cv2(image)
        cv2.imwrite(filename, cv2_img)

    def save_depth_image(self, image, filename):
        """
        Save a normalized (easier to visualize) version
        of a depth image into a file.
        """
        # Duc's smart undocummented code
        im_array = self.convert_depth_to_cv2(image)
        min_distance, max_distance = np.min(im_array), np.max(im_array)
        im_array = im_array * 1.0
        im_array = (im_array < max_distance) * im_array
        im_array = (im_array - min_distance) / max_distance * 255.0
        im_array = (im_array >= 0) * im_array

        cv2.imwrite(filename, im_array)

    def load_from_file(self, file_path, cv2_imread_mode=None):
        """
        Load image from a file.
        :param file_path str: Path to the image file.
        :param cv2_imread_mode int: cv2.IMREAD_ mode, modes are:
            cv2.IMREAD_ANYCOLOR 4             cv2.IMREAD_REDUCED_COLOR_4 33
            cv2.IMREAD_ANYDEPTH 2             cv2.IMREAD_REDUCED_COLOR_8 65
            cv2.IMREAD_COLOR 1                cv2.IMREAD_REDUCED_GRAYSCALE_2 16
            cv2.IMREAD_GRAYSCALE 0            cv2.IMREAD_REDUCED_GRAYSCALE_4 32
            cv2.IMREAD_IGNORE_ORIENTATION 128 cv2.IMREAD_REDUCED_GRAYSCALE_8 64
            cv2.IMREAD_LOAD_GDAL 8            cv2.IMREAD_UNCHANGED -1
            cv2.IMREAD_REDUCED_COLOR_2 17
        """
        if cv2_imread_mode is not None:
            img = cv2.imread(file_path, cv2_imread_mode)
        img = cv2.imread(file_path)
        if img is None:
            raise RuntimeError("No image found to load at " + str(file_path))
        return img


if __name__ == '__main__':
    from sensor_msgs.msg import Image, CompressedImage
    it = ImageTools()
    from cPickle import load
    img = load(open('rgb_image.pickle', 'r'))
    cv2_img = it.convert_ros_msg_to_cv2(img)
    print(type(cv2_img))
    ros_img = it.convert_cv2_to_ros_msg(cv2_img)
    print(type(ros_img))
    ros_compressed2 = it.convert_cv2_to_ros_compressed_msg(cv2_img)
    print(type(ros_compressed2))
    compressed_ros_img = it.convert_ros_msg_to_ros_compressed_msg(ros_img)
    print(type(compressed_ros_img))
    ros_img2 = it.convert_ros_compressed_msg_to_ros_msg(compressed_ros_img)
    print(type(ros_img2))
    cv2_2 = it.convert_ros_compressed_to_cv2(compressed_ros_img)
    print(type(cv2_2))

    cv2_3 = it.convert_to_cv2(cv2_img)
    cv2_4 = it.convert_to_cv2(ros_img)
    cv2_5 = it.convert_to_cv2(compressed_ros_img)

    ros_3 = it.convert_to_ros_msg(cv2_img)
    ros_4 = it.convert_to_ros_msg(ros_img)
    ros_5 = it.convert_to_ros_msg(compressed_ros_img)

    rosc_3 = it.convert_to_ros_compressed_msg(cv2_img)
    rosc_4 = it.convert_to_ros_compressed_msg(ros_img)
    rosc_5 = it.convert_to_ros_compressed_msg(compressed_ros_img)

    depthcompimg = load(open('depth_compressed_image.pickle', 'r'))
    depth_cv2 = it.convert_depth_to_cv2(depthcompimg)
    depth_ros = it.convert_depth_to_ros_msg(depthcompimg)
    depth_ros_comp = it.convert_depth_to_ros_compressed_msg(depthcompimg)
    it.save_image(depth_cv2, 'depth_comp_cv2.jpg')
    it.save_depth_image(depth_cv2, 'depth_comp_normalized_cv2.jpg')

    depthimg = load(open('depth_image.pickle', 'r'))
    depthimg_cv2 = it.convert_depth_to_cv2(depthimg)
    depthimg_ros = it.convert_depth_to_ros_msg(depthimg)
    depthimg_ros_comp = it.convert_depth_to_ros_compressed_msg(depthimg)
    it.save_image(depth_cv2, 'depth_cv2.jpg')
    it.save_depth_image(depth_cv2, 'depth_normalized_cv2.jpg')

    it.save_image(cv2_img, 'cv2_image.jpg')
    it.save_image(ros_img, 'ros_image.jpg')
    it.save_image(compressed_ros_img, 'compressed_ros_image.jpg')

    it.display_image(cv2_img)

Do you have 32FC1 conversion functions? 32FC1; compressedDepth png is of particular interest.

Hey @lucasw , I don't have 32FC1 sorry. The robot that required this work only published in 16UC1.

I think I figured it out here https://answers.ros.org/question/407659/how-to-generate-format-32fc1-compresseddepth-png-image-in-python-from-float-array/ - and it turned out it isn't actually 32fc1 despite the format name, just uint16 with a scale and offset (and the depth values get inverted)

That's an odd format! Thanks for figuring it out. I'm going to quote here your reply there just for the sake of conservation of information:

Here is what I came up with- I haven't yet checked to make sure the depth values when decompressed are correct, they just seemed reasonable as seen in rqt:

import cv2
import numpy as np
from sensor_msgs.msg import CompressedImage


def encode_compressed_depth_image_msg(depth_image: np.array,
                                      depth_min=1.0, depth_max=10.0,
                                      depth_quantization=100.0):
    depth_image[depth_image > depth_max] = np.nan
    depth_image[depth_image < depth_min] = np.nan
    depth_z0 = depth_quantization
    depth_quant_a = depth_z0 * (depth_z0 + 1.0)
    depth_quant_b = 1.0 - depth_quant_a / depth_max
    inv_depth = (depth_quant_a / depth_image + depth_quant_b).astype(np.uint16)
    depth_encoded = cv2.imencode(".png", inv_depth)[1]
    header = np.array([0.0, depth_quant_a, depth_quant_b], np.float32)

    compressed_depth_msg = CompressedImage()
    compressed_depth_msg.format = "32FC1; compressedDepth png"
    compressed_depth_msg.data = header.tobytes() + depth_encoded.tobytes()
    return compressed_depth_msg

The payload of the CompressedDepth 32FC1; compressedDepth png is 12 header bytes which contains the conversion scaling numbers quant a & b, then an encoded 16-bit grayscale png follows (if you lop off those first 12 bytes and save the rest of the bytes to a file you can display it in a png viewer and see the depth values). The scaling numbers fixed above but could be changed per-frame if there was a reason to, by manipulating depth_max and depth_quantization (or refactor so depth_quant_a and b are set through some other means).

@awesomebytes thank you very much for publishing these conversion tools. For 2D RGB images, they work really well.

@Mechazo11 You are welcome! I'm very happy to know my snippets help people! :)