diff --git a/modules/videoio/src/cap_gstreamer.cpp b/modules/videoio/src/cap_gstreamer.cpp
index e6edccdf5..1e2339eac 100644
--- a/modules/videoio/src/cap_gstreamer.cpp
+++ b/modules/videoio/src/cap_gstreamer.cpp
@@ -292,18 +292,22 @@ IplImage * CvCapture_GStreamer::retrieveFrame(int)
         const gchar* name = gst_structure_get_name(structure);
         const gchar* format = gst_structure_get_string(structure, "format");
 
-        if (!name || !format)
+        if (!name)
             return 0;
 
         // we support 3 types of data:
         //     video/x-raw, format=BGR   -> 8bit, 3 channels
         //     video/x-raw, format=GRAY8 -> 8bit, 1 channel
         //     video/x-bayer             -> 8bit, 1 channel
+        //     image/jpeg                -> 8bit, mjpeg: buffer_size x 1 x 1
         // bayer data is never decoded, the user is responsible for that
         // everything is 8 bit, so we just test the caps for bit depth
 
         if (strcasecmp(name, "video/x-raw") == 0)
         {
+            if (!format)
+                return 0;
+
             if (strcasecmp(format, "BGR") == 0) {
                 depth = 3;
             }
@@ -314,6 +318,9 @@ IplImage * CvCapture_GStreamer::retrieveFrame(int)
         else if (strcasecmp(name, "video/x-bayer") == 0)
         {
             depth = 1;
+        } else if(strcasecmp(name, "image/jpeg") == 0) {
+            depth = 1;
+            // the correct size will be set once the first frame arrives
         }
 #endif
         if (depth > 0) {
@@ -334,6 +341,13 @@ IplImage * CvCapture_GStreamer::retrieveFrame(int)
     // info.data ptr is valid until next grabFrame where the associated sample is unref'd
     GstMapInfo info = GstMapInfo();
     gboolean success = gst_buffer_map(buffer,&info, (GstMapFlags)GST_MAP_READ);
+
+    // with MJPEG streams frame size can change arbitrarily
+    if(int(info.size) != frame->imageSize) {
+        cvReleaseImageHeader(&frame);
+        frame = cvCreateImageHeader(cvSize(info.size, 1), IPL_DEPTH_8U, 1);
+    }
+
     if (!success){
         //something weird went wrong here. abort. abort.
         //fprintf(stderr,"GStreamer: unable to map buffer");
@@ -797,7 +811,7 @@ bool CvCapture_GStreamer::open( int type, const char* filename )
                                NULL);
 #else
     // support 1 and 3 channel 8 bit data, as well as bayer (also  1 channel, 8bit)
-    caps = gst_caps_from_string("video/x-raw, format=(string){BGR, GRAY8}; video/x-bayer,format=(string){rggb,bggr,grbg,gbrg}");
+    caps = gst_caps_from_string("video/x-raw, format=(string){BGR, GRAY8}; video/x-bayer,format=(string){rggb,bggr,grbg,gbrg}; image/jpeg");
 #endif
     gst_app_sink_set_caps(GST_APP_SINK(sink), caps);
     gst_caps_unref(caps);