Import opencv mat in python (compressed with c#)

Question

I've problems decoding/unzipping a string in python which was orginially compressed using C# with "GZipStream".

The "opencv mat" was converted to json with:

public override void WriteJson(JsonWriter writer, Mat value, JsonSerializer serializer)
{
    byte[] matBytes = null;
    Cv2.ImEncode(".png", value, out matBytes);

    // Convert byte[] to Base64 String
    string matBase64 = Convert.ToBase64String(matBytes);

    var resCompressed = Utils.StringCompressor.CompressString(matBase64);
    writer.WriteValue(resCompressed);
}

An example string "jsontoken" to uncompress is:

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

The according C# code to load this data is:

private static void CreateDataFromJsonZip(ref ThresholdMask mask, JObject jsonObj, string maskName)
{
    // get JSON result objects into a list
    JToken jsonToken = jsonObj["Selection"]; # jsonToken

    var res = Utils.StringCompressor.DecompressString(jsonToken.ToString());
    byte[] data = Convert.FromBase64String(res);
    mask.Selection = Cv2.ImDecode(data, ImreadModes.Unchanged);

My attempt to read the data in python is:

jsonToken = r'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'

string = jsonToken
original = base64.b64decode(string)
convert_to_np = np.frombuffer(original, dtype=np.uint8)
print(convert_to_np)
len(convert_to_np)
img = cv2.imdecode(convert_to_np, cv2.IMREAD_UNCHANGED)

Unfortunately, the last step with imdecode does not work, the result is None. Any hints?

"json"? there's no json there. what does `Utils.StringCompressor.CompressString` do? and that data isn't a valid PNG file either. what did you do in the C# part of your code? why does this code talk about "strings" when it's slinging binary data around? the only actual strings here are base64-encoded data. everything else MUST be binary-safe. python has no issue with that, nor does OpenCV. the stuff you do in C#, does that do the right thing? — Christoph Rackwitz, Sep 21 '21 at 21:02

score 1 · Accepted Answer · answered Sep 21 '21 at 23:13

This will do exactly what you asked for, following the StringCompressor.DecompressString code from this SO answer (the decompressed length is stored in the first 4 bytes):

original = base64.b64decode(gzip.decompress(base64.b64decode(jsonToken)[4:]))

However, I'd recommend simplifying your C# implementation. Starting from the raw OpenCV mat, you:

encode it as a PNG image
convert that binary data to base64 string
compress the text string with gzip
convert to base64 again
store it to the json file

Step 2 is completely unnecessary, and step 3 is redundant, since the PNG encoding process already compresses the image data. Adding another layer of compression won't help much, and may even make it worse.

I would just convert to a PNG, encode with base64, and write that to the json file. Then your original Python code would work as expected.

Import opencv mat in python (compressed with c#)

1 Answers1