Merge pull request #25584 from dkurt:videocapture_from_buffer

Open VideoCapture from data stream #25584

### Pull Request Readiness Checklist

Add VideoCapture option to read a raw binary video data from `std::streambuf`.

There are multiple motivations:
1. Avoid disk file creation in case of video already in memory (received by network or from database).
2. Streaming mode. Frames decoding starts during sequential file transfer by chunks.

Suppoted backends:
* FFmpeg
* MSMF (no streaming mode)

Supporter interfaces:
* C++ (std::streambuf)
* Python (io.BufferedIOBase)

resolves https://github.com/opencv/opencv/issues/24400

- [x] test h264
- [x]  test IP camera like approach with no metadata but key frame only?
- [x] C API plugin

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [x] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
      Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake

modules/videoio/misc/python/pyopencv_videoio.hpp

@@ -31,4 +31,114 @@ template<> bool pyopencv_to(PyObject* obj, cv::VideoCapture& stream, const ArgIn
     return true;
 }
 
+class PythonStreamReader : public cv::IStreamReader
+{
+public:
+    PythonStreamReader(PyObject* _obj = nullptr) : obj(_obj)
+    {
+        if (obj)
+            Py_INCREF(obj);
+    }
+
+    ~PythonStreamReader()
+    {
+        if (obj)
+            Py_DECREF(obj);
+    }
+
+    long long read(char* buffer, long long size) CV_OVERRIDE
+    {
+        if (!obj)
+            return 0;
+
+        PyObject* ioBase = reinterpret_cast<PyObject*>(obj);
+
+        PyGILState_STATE gstate;
+        gstate = PyGILState_Ensure();
+
+        PyObject* py_size = pyopencv_from(static_cast<int>(size));
+
+        PyObject* res = PyObject_CallMethodObjArgs(ioBase, PyString_FromString("read"), py_size, NULL);
+        bool hasPyReadError = PyErr_Occurred() != nullptr;
+        char* src = PyBytes_AsString(res);
+        size_t len = static_cast<size_t>(PyBytes_Size(res));
+        bool hasPyBytesError = PyErr_Occurred() != nullptr;
+        if (src && len <= static_cast<size_t>(size))
+        {
+            std::memcpy(buffer, src, len);
+        }
+        Py_DECREF(res);
+        Py_DECREF(py_size);
+
+        PyGILState_Release(gstate);
+
+        if (hasPyReadError)
+            CV_Error(cv::Error::StsError, "Python .read() call error");
+        if (hasPyBytesError)
+            CV_Error(cv::Error::StsError, "Python buffer access error");
+
+        CV_CheckLE(len, static_cast<size_t>(size), "Stream chunk size should be less or equal than requested size");
+
+        return len;
+    }
+
+    long long seek(long long offset, int way) CV_OVERRIDE
+    {
+        if (!obj)
+            return 0;
+
+        PyObject* ioBase = reinterpret_cast<PyObject*>(obj);
+
+        PyGILState_STATE gstate;
+        gstate = PyGILState_Ensure();
+
+        PyObject* py_offset = pyopencv_from(static_cast<int>(offset));
+        PyObject* py_whence = pyopencv_from(way);
+
+        PyObject* res = PyObject_CallMethodObjArgs(ioBase, PyString_FromString("seek"), py_offset, py_whence, NULL);
+        bool hasPySeekError = PyErr_Occurred() != nullptr;
+        long long pos = PyLong_AsLongLong(res);
+        bool hasPyConvertError = PyErr_Occurred() != nullptr;
+        Py_DECREF(res);
+        Py_DECREF(py_offset);
+        Py_DECREF(py_whence);
+
+        PyGILState_Release(gstate);
+
+        if (hasPySeekError)
+            CV_Error(cv::Error::StsError, "Python .seek() call error");
+        if (hasPyConvertError)
+            CV_Error(cv::Error::StsError, "Python .seek() result => long long conversion error");
+        return pos;
+    }
+
+private:
+    PyObject* obj;
+};
+
+template<>
+bool pyopencv_to(PyObject* obj, Ptr<cv::IStreamReader>& p, const ArgInfo&)
+{
+    if (!obj)
+        return false;
+
+    PyObject* ioModule = PyImport_ImportModule("io");
+    PyObject* type = PyObject_GetAttrString(ioModule, "BufferedIOBase");
+    Py_DECREF(ioModule);
+    bool isValidPyType = PyObject_IsInstance(obj, type) == 1;
+    Py_DECREF(type);
+
+    if (!isValidPyType)
+    {
+        PyErr_SetString(PyExc_TypeError, "Input stream should be derived from io.BufferedIOBase");
+        return false;
+    }
+
+    if (!PyErr_Occurred()) {
+        p = makePtr<PythonStreamReader>(obj);
+        return true;
+    }
+    return false;
+}
+
 #endif // HAVE_OPENCV_VIDEOIO

modules/videoio/misc/python/test/test_videoio.py

@@ -3,6 +3,8 @@ from __future__ import print_function
 
 import numpy as np
 import cv2 as cv
+import io
+import sys
 
 from tests_common import NewOpenCVTests
 
@@ -21,5 +23,69 @@ class Bindings(NewOpenCVTests):
         for backend in backends:
             self.check_name(cv.videoio_registry.getBackendName(backend))
 
+    def test_capture_stream_file(self):
+        if sys.version_info[0] < 3:
+            raise self.skipTest('Python 3.x required')
+
+        api_pref = None
+        for backend in cv.videoio_registry.getStreamBufferedBackends():
+            if not cv.videoio_registry.hasBackend(backend):
+                continue
+            if not cv.videoio_registry.isBackendBuiltIn(backend):
+                _, abi, api = cv.videoio_registry.getStreamBufferedBackendPluginVersion(backend)
+                if (abi < 1 or (abi == 1 and api < 2)):
+                    continue
+            api_pref = backend
+            break
+
+        if not api_pref:
+            raise self.skipTest("No available backends")
+
+        with open(self.find_file("cv/video/768x576.avi"), "rb") as f:
+            cap = cv.VideoCapture(f, api_pref, [])
+            self.assertTrue(cap.isOpened())
+            hasFrame, frame = cap.read()
+            self.assertTrue(hasFrame)
+            self.assertEqual(frame.shape, (576, 768, 3))
+
+    def test_capture_stream_buffer(self):
+        if sys.version_info[0] < 3:
+            raise self.skipTest('Python 3.x required')
+
+        api_pref = None
+        for backend in cv.videoio_registry.getStreamBufferedBackends():
+            if not cv.videoio_registry.hasBackend(backend):
+                continue
+            if not cv.videoio_registry.isBackendBuiltIn(backend):
+                _, abi, api = cv.videoio_registry.getStreamBufferedBackendPluginVersion(backend)
+                if (abi < 1 or (abi == 1 and api < 2)):
+                    continue
+            api_pref = backend
+            break
+
+        if not api_pref:
+            raise self.skipTest("No available backends")
+
+        class BufferStream(io.BufferedIOBase):
+            def __init__(self, filepath):
+                self.f = open(filepath, "rb")
+
+            def read(self, size=-1):
+                return self.f.read(size)
+
+            def seek(self, offset, whence):
+                return self.f.seek(offset, whence)
+
+            def __del__(self):
+                self.f.close()
+
+        stream = BufferStream(self.find_file("cv/video/768x576.avi"))
+
+        cap = cv.VideoCapture(stream, api_pref, [])
+        self.assertTrue(cap.isOpened())
+        hasFrame, frame = cap.read()
+        self.assertTrue(hasFrame)
+        self.assertEqual(frame.shape, (576, 768, 3))
+
 if __name__ == '__main__':
     NewOpenCVTests.bootstrap()