diff --git a/gst/rtp/gstrtpjpegdepay.c b/gst/rtp/gstrtpjpegdepay.c
index f10bf464f7..0dbc736345 100644
--- a/gst/rtp/gstrtpjpegdepay.c
+++ b/gst/rtp/gstrtpjpegdepay.c
@@ -37,6 +37,11 @@ GST_STATIC_PAD_TEMPLATE ("src",
     GST_PAD_SRC,
     GST_PAD_ALWAYS,
     GST_STATIC_CAPS ("image/jpeg")
+        /* optional SDP attributes */
+        /*
+         * "width = (int) 0, "
+         * "height = (int) 0, "
+         */
     );
 
 static GstStaticPadTemplate gst_rtp_jpeg_depay_sink_template =
@@ -45,13 +50,15 @@ static GstStaticPadTemplate gst_rtp_jpeg_depay_sink_template =
     GST_PAD_ALWAYS,
     GST_STATIC_CAPS ("application/x-rtp, "
         "media = (string) \"video\", "
-        "clock-rate = (int) 90000, " "encoding-name = (string) \"JPEG\"; "
+        "clock-rate = (int) 90000, "
+        "encoding-name = (string) \"JPEG\"; "
         /* optional SDP attributes */
         /*
+         * "width = (int) 0, "
+         * "height = (int) 0, "
          * "a-framerate = (string) 0.00, "
          * "x-framerate = (string) 0.00, "
-         * "a-framesize = (string) 1234-1234, "
-         * "x-dimensions = (string) \"1234,1234\", "
+         * "x-dimensions = (string) "0\,0", "
          */
         "application/x-rtp, "
         "media = (string) \"video\", "
@@ -59,10 +66,11 @@ static GstStaticPadTemplate gst_rtp_jpeg_depay_sink_template =
         "clock-rate = (int) 90000"
         /* optional SDP attributes */
         /*
+         * "width = (int) 0, "
+         * "height = (int) 0, "
          * "a-framerate = (string) 0.00, "
          * "x-framerate = (string) 0.00, "
-         * "a-framesize = (string) 1234-1234, "
-         * "x-dimensions = (string) \"1234,1234\""
+         * "x-dimensions = (string) "0\,0", "
          */
     )
     );
@@ -430,6 +438,7 @@ gst_rtp_jpeg_depay_setcaps (GstRTPBaseDepayload * depayload, GstCaps * caps)
   GstStructure *structure;
   gint clock_rate;
   const gchar *media_attr;
+  gint width = 0, height = 0;
 
   rtpjpegdepay = GST_RTP_JPEG_DEPAY (depayload);
 
@@ -450,21 +459,17 @@ gst_rtp_jpeg_depay_setcaps (GstRTPBaseDepayload * depayload, GstCaps * caps)
 
   /* check for optional SDP attributes */
   if ((media_attr = gst_structure_get_string (structure, "x-dimensions"))) {
-    gint w, h;
-
-    if (sscanf (media_attr, "%d,%d", &w, &h) == 2) {
-      rtpjpegdepay->media_width = w;
-      rtpjpegdepay->media_height = h;
+    if (sscanf (media_attr, "%d,%d", &width, &height) != 2 || width <= 0 ||
+        height <= 0) {
+      goto invalid_dimension;
     }
   }
 
-  if ((media_attr = gst_structure_get_string (structure, "a-framesize"))) {
-    gint w, h;
-
-    if (sscanf (media_attr, "%d-%d", &w, &h) == 2) {
-      rtpjpegdepay->media_width = w;
-      rtpjpegdepay->media_height = h;
-    }
+  if (gst_structure_get_int (structure, "width", &width) && width <= 0) {
+    goto invalid_dimension;
+  }
+  if (gst_structure_get_int (structure, "height", &height) && height <= 0) {
+    goto invalid_dimension;
   }
 
   /* try to get a framerate */
@@ -494,7 +499,16 @@ gst_rtp_jpeg_depay_setcaps (GstRTPBaseDepayload * depayload, GstCaps * caps)
     g_free (s);
   }
 
+  rtpjpegdepay->media_width = width;
+  rtpjpegdepay->media_height = height;
+
   return TRUE;
+
+invalid_dimension:
+  {
+    GST_ERROR_OBJECT (rtpjpegdepay, "invalid width/height from caps");
+    return FALSE;
+  }
 }
 
 static GstBuffer *
@@ -631,8 +645,13 @@ gst_rtp_jpeg_depay_process (GstRTPBaseDepayload * depayload, GstBuffer * buf)
 
       outcaps =
           gst_caps_new_simple ("image/jpeg", "framerate", GST_TYPE_FRACTION,
-          rtpjpegdepay->frate_num, rtpjpegdepay->frate_denom, "width",
-          G_TYPE_INT, width, "height", G_TYPE_INT, height, NULL);
+          rtpjpegdepay->frate_num, rtpjpegdepay->frate_denom, NULL);
+
+      if (width > 0 && height > 0) {
+        gst_caps_set_simple (outcaps, "width", G_TYPE_INT, width, "height",
+            G_TYPE_INT, height, NULL);
+      }
+
       gst_pad_set_caps (depayload->srcpad, outcaps);
       gst_caps_unref (outcaps);
 
diff --git a/gst/rtp/gstrtpjpegpay.c b/gst/rtp/gstrtpjpegpay.c
index 859d16b16c..bea26e6170 100644
--- a/gst/rtp/gstrtpjpegpay.c
+++ b/gst/rtp/gstrtpjpegpay.c
@@ -45,7 +45,12 @@ static GstStaticPadTemplate gst_rtp_jpeg_pay_sink_template =
     GST_STATIC_PAD_TEMPLATE ("sink",
     GST_PAD_SINK,
     GST_PAD_ALWAYS,
-    GST_STATIC_CAPS ("image/jpeg; " "video/x-jpeg")
+    GST_STATIC_CAPS ("image/jpeg, "
+        "  width = (int) [ 1, MAX ], "
+        "  height = (int) [ 1, MAX ]; "
+        "  video/x-jpeg, "
+        "  width = (int) [ 1, MAX ], "
+        "  height = (int) [ 1, MAX ]")
     );
 
 static GstStaticPadTemplate gst_rtp_jpeg_pay_src_template =
@@ -57,8 +62,8 @@ GST_STATIC_PAD_TEMPLATE ("src",
         "  payload = (int) 26 ,        "
         "  clock-rate = (int) 90000,   "
         "  encoding-name = (string) \"JPEG\", "
-        "  width = (int) [ 1, 65536 ], "
-        "  height = (int) [ 1, 65536 ]")
+        "  width = (int) [ 1, MAX ], "
+        "  height = (int) [ 1, MAX ]")
     );
 
 GST_DEBUG_CATEGORY_STATIC (rtpjpegpay_debug);
@@ -293,11 +298,9 @@ gst_rtp_jpeg_pay_setcaps (GstRTPBasePayload * basepayload, GstCaps * caps)
   GstStructure *caps_structure = gst_caps_get_structure (caps, 0);
   GstRtpJPEGPay *pay;
   gboolean res;
-  gint width = -1, height = -1;
+  gint width, height;
   gint num = 0, denom;
   gchar *rate = NULL;
-  gchar *dim = NULL;
-  gchar *size;
 
   pay = GST_RTP_JPEG_PAY (basepayload);
 
@@ -317,6 +320,8 @@ gst_rtp_jpeg_pay_setcaps (GstRTPBasePayload * basepayload, GstCaps * caps)
   }
 
   if (height > 2040 || width > 2040) {
+    GST_DEBUG_OBJECT (pay,
+        "width or height > 2040, need to rely on caps instead of RTP header");
     pay->height = 0;
     pay->width = 0;
   } else {
@@ -333,34 +338,17 @@ gst_rtp_jpeg_pay_setcaps (GstRTPBasePayload * basepayload, GstCaps * caps)
     rate = g_strdup_printf("%f", framerate);
   }
 
-  size = g_strdup_printf("%d-%d", width, height);
-
-  if (pay->width == 0) {
-    GST_DEBUG_OBJECT (pay,
-        "width or height are greater than 2040, adding x-dimensions to caps");
-    dim = g_strdup_printf ("%d,%d", width, height);
+  if (rate != NULL) {
+    res = gst_rtp_base_payload_set_outcaps (basepayload, "width", G_TYPE_INT,
+        width, "height", G_TYPE_INT, height, "a-framerate", G_TYPE_STRING,
+        rate, NULL);
+  } else if (rate == NULL) {
+    res = gst_rtp_base_payload_set_outcaps (basepayload, "width",
+        G_TYPE_INT, width, "height", G_TYPE_INT, height, NULL);
   }
 
-  if (rate != NULL && dim != NULL) {
-    res = gst_rtp_base_payload_set_outcaps (basepayload, "a-framerate",
-        G_TYPE_STRING, rate, "a-framesize", G_TYPE_STRING, size,
-        "x-dimensions", G_TYPE_STRING, dim, NULL);
-  } else if (rate != NULL && dim == NULL) {
-    res = gst_rtp_base_payload_set_outcaps (basepayload, "a-framerate",
-        G_TYPE_STRING, rate, "a-framesize", G_TYPE_STRING, size, NULL);
-  } else if (rate == NULL && dim != NULL) {
-    res = gst_rtp_base_payload_set_outcaps (basepayload, "x-dimensions",
-        G_TYPE_STRING, dim, "a-framesize", G_TYPE_STRING, size, NULL);
-  } else {
-    res = gst_rtp_base_payload_set_outcaps (basepayload, "a-framesize",
-        G_TYPE_STRING, size, NULL);
-  }
-
-  if (dim != NULL)
-    g_free (dim);
   if (rate != NULL)
     g_free (rate);
-  g_free (size);
 
   return res;