Multi-Threaded Image Processing using Qt Camera - Revision history

Hamishwillee: Text replace - "`" to ""`


2012-10-11T01:17:58Z
Text replace - "<code cpp>" to "<code cpp-qt>"
Show changes



Hamishwillee: Text replace - "Category:MeeGo" to "Category:MeeGo Harmattan"
2012-06-13T10:55:57Z
Text replace - "Category:MeeGo" to "Category:MeeGo Harmattan"

			
			
			
			
		
		← Older revision
		Revision as of 10:55, 13 June 2012
		
Line 1:
Line 1:
− [[Category:Camera]][[Category:Performance]][[Category:Symbian]][[Category:Code Examples]][[Category:Imaging]][[Category:Qt]][[Category:Qt Mobility]][[Category:MeeGo]][[Category:Multimedia]] + [[Category:Camera]][[Category:Performance]][[Category:Symbian]][[Category:Code Examples]][[Category:Imaging]][[Category:Qt]][[Category:Qt Mobility]][[Category:MeeGo Harmattan]][[Category:Multimedia]]
  {{Abstract|This article shows how to apply live image processing effects to the camera viewfinder using multi-threading and capture full resolution snapshots.}}   {{Abstract|This article shows how to apply live image processing effects to the camera viewfinder using multi-threading and capture full resolution snapshots.}}
  {{SeeAlso|   {{SeeAlso|




Hamishwillee: Hamishwillee - Fix typo
2012-05-31T01:01:50Z
Hamishwillee - Fix typo

			
			
			
			
		
		← Older revision
		Revision as of 01:01, 31 May 2012
		
Line 1:
Line 1:
  [[Category:Camera]][[Category:Performance]][[Category:Symbian]][[Category:Code Examples]][[Category:Imaging]][[Category:Qt]][[Category:Qt Mobility]][[Category:MeeGo]][[Category:Multimedia]]   [[Category:Camera]][[Category:Performance]][[Category:Symbian]][[Category:Code Examples]][[Category:Imaging]][[Category:Qt]][[Category:Qt Mobility]][[Category:MeeGo]][[Category:Multimedia]]
  {{Abstract|This article shows how to apply live image processing effects to the camera viewfinder using multi-threading and capture full resolution snapshots.}}   {{Abstract|This article shows how to apply live image processing effects to the camera viewfinder using multi-threading and capture full resolution snapshots.}}
− {{SeeAlso|[[How to add multithread support to your Qt application]] + {{SeeAlso|
  + * [[How to add multithread support to your Qt application]]
  * [http://qt-project.org/doc/qt-4.8/thread-basics.html Threading basic] (Qt reference)   * [http://qt-project.org/doc/qt-4.8/thread-basics.html Threading basic] (Qt reference)
  * [http://qt-project.org/doc/qt-4.8/threads.html Thread support in Qt] (Qt reference)}}   * [http://qt-project.org/doc/qt-4.8/threads.html Thread support in Qt] (Qt reference)}}




Hamishwillee: Hamishwillee - Add see also to similar articles and reference material on topic
2012-05-31T01:01:08Z
Hamishwillee - Add see also to similar articles and reference material on topic

			
			
			
			
		
		← Older revision
		Revision as of 01:01, 31 May 2012
		
Line 1:
Line 1:
  [[Category:Camera]][[Category:Performance]][[Category:Symbian]][[Category:Code Examples]][[Category:Imaging]][[Category:Qt]][[Category:Qt Mobility]][[Category:MeeGo]][[Category:Multimedia]]   [[Category:Camera]][[Category:Performance]][[Category:Symbian]][[Category:Code Examples]][[Category:Imaging]][[Category:Qt]][[Category:Qt Mobility]][[Category:MeeGo]][[Category:Multimedia]]
  {{Abstract|This article shows how to apply live image processing effects to the camera viewfinder using multi-threading and capture full resolution snapshots.}}   {{Abstract|This article shows how to apply live image processing effects to the camera viewfinder using multi-threading and capture full resolution snapshots.}}
−   + {{SeeAlso|[[How to add multithread support to your Qt application]]
  + * [http://qt-project.org/doc/qt-4.8/thread-basics.html Threading basic] (Qt reference)
  + * [http://qt-project.org/doc/qt-4.8/threads.html Thread support in Qt] (Qt reference)}}
  {{Note|This is an entry in the [[PureView Imaging Competition 2012Q2]]}}   {{Note|This is an entry in the [[PureView Imaging Competition 2012Q2]]}}
−  
  {{ArticleMetaData <!-- v1.2 -->   {{ArticleMetaData <!-- v1.2 -->
  |sourcecode= [[Media:Fipcamera.zip]]   |sourcecode= [[Media:Fipcamera.zip]]




Meh.at: Meh.at - /* When should multi-threading be used? */
2012-05-22T05:18:56Z
Meh.at - - When should multi-threading be used?

			
			
			
			
		
		← Older revision
		Revision as of 05:18, 22 May 2012
		
Line 616:
Line 616:
  == When should multi-threading be used? ==   == When should multi-threading be used? ==
  Multi-threading incorporates additional processing and memory costs for thread management, inter-thread communication, locking, and data copying. As a rule-of-thumb multi-threading should be used if a task (e.g. image processing) is expected to take longer than 1 second (see Gnome guidelines for desktop [http://developer.gnome.org/hig-book/3.2/feedback-response-times.html.en here]; users of mobile devices accept only very short response times!).   Multi-threading incorporates additional processing and memory costs for thread management, inter-thread communication, locking, and data copying. As a rule-of-thumb multi-threading should be used if a task (e.g. image processing) is expected to take longer than 1 second (see Gnome guidelines for desktop [http://developer.gnome.org/hig-book/3.2/feedback-response-times.html.en here]; users of mobile devices accept only very short response times!).
  + 
  + This article focuses on multi-threading for user interface responsiveness by processing live data in a worker thread. But multi-threading is also useful when multiple tasks can be run in parallel, as for instance in the [[QHdrCamera_component_for_High_Dynamic_Range_Imaging|QHdrCamera]] project where image processing starts during capturing of remaining pictures which decreases processing time significantly.
    
  == Summary ==   == Summary ==
  This article presents a brief overview on how to apply near-real time image processing effects to a camera viewfinder using QML and how to capture full-resolution snapshots. It outlines how image processing can be moved to a worker thread and how to handle concurrent access using double buffering and mutex. Finally, optimizations are discussed.   This article presents a brief overview on how to apply near-real time image processing effects to a camera viewfinder using QML and how to capture full-resolution snapshots. It outlines how image processing can be moved to a worker thread and how to handle concurrent access using double buffering and mutex. Finally, optimizations are discussed.




Meh.at: Meh.at -
2012-05-18T19:47:31Z
Meh.at -

			
			
			
			
		
		← Older revision
		Revision as of 19:47, 18 May 2012
		
Line 615:
Line 615:
    
  == When should multi-threading be used? ==   == When should multi-threading be used? ==
− Multi-threading incorporates additional processing and memory costs for thread management, inter-thread communication, locking, and data copying. As a rule-of-thumb multi-threading should be used if a task (e.g. image processing) is expected to take longer than 1 second (see Gnome guidelines for Desktop [http://developer.gnome.org/hig-book/3.2/feedback-response-times.html.en here]; users of mobile devices accept only very short response times!). + Multi-threading incorporates additional processing and memory costs for thread management, inter-thread communication, locking, and data copying. As a rule-of-thumb multi-threading should be used if a task (e.g. image processing) is expected to take longer than 1 second (see Gnome guidelines for desktop [http://developer.gnome.org/hig-book/3.2/feedback-response-times.html.en here]; users of mobile devices accept only very short response times!).
    
  == Summary ==   == Summary ==
  This article presents a brief overview on how to apply near-real time image processing effects to a camera viewfinder using QML and how to capture full-resolution snapshots. It outlines how image processing can be moved to a worker thread and how to handle concurrent access using double buffering and mutex. Finally, optimizations are discussed.   This article presents a brief overview on how to apply near-real time image processing effects to a camera viewfinder using QML and how to capture full-resolution snapshots. It outlines how image processing can be moved to a worker thread and how to handle concurrent access using double buffering and mutex. Finally, optimizations are discussed.




Meh.at: Meh.at - /* Discussion about hardware acceleration */
2012-05-18T19:45:20Z
Meh.at - - Discussion about hardware acceleration

			
			
			
			
		
		← Older revision
		Revision as of 19:45, 18 May 2012
		
Line 612:
Line 612:
  It can be seen that ARM-based conversion is the fastest and sums up to 10.69ms for conversion and drawing. The CPU-based conversion requires almost double the time and sums up to 14.78ms. OpenGL ES 2 shader based drawing and conversion requires 7.82ms but the time for upload is very high and requires 18.03 which sums up to a total of 25.85ms.   It can be seen that ARM-based conversion is the fastest and sums up to 10.69ms for conversion and drawing. The CPU-based conversion requires almost double the time and sums up to 14.78ms. OpenGL ES 2 shader based drawing and conversion requires 7.82ms but the time for upload is very high and requires 18.03 which sums up to a total of 25.85ms.
    
− From a performance perspective the best means for data conversion and drawing is ARM conversion with "normal" drawing. But with QML 1.2 only OpenGL-based drawing is supported. Thus, if QML is used, the best method is to do conversion and drawing with OpenGL ES 2 shaders. Finally, it should be noted that all methods are fine for live video preview because with a frame rate of 30fps the available frame time is 33.33ms. + From a performance perspective the best means for data conversion and drawing is ARM-based conversion with "normal" drawing. But with QML 1.2 on the Nokia N9 only OpenGL-based drawing is supported. Thus, if QML is used, the best method is to do conversion and drawing with OpenGL ES 2 shaders. Finally, it should be noted that all methods are fine for live video preview because with a frame rate of 30fps the available frame time is 33.33ms.
    
  == When should multi-threading be used? ==   == When should multi-threading be used? ==




Meh.at: Meh.at - /* Discussion about hardware acceleration */
2012-05-18T19:43:42Z
Meh.at - - Discussion about hardware acceleration

			
			
			
			
		
		← Older revision
		Revision as of 19:43, 18 May 2012
		
Line 611:
Line 611:
    
  It can be seen that ARM-based conversion is the fastest and sums up to 10.69ms for conversion and drawing. The CPU-based conversion requires almost double the time and sums up to 14.78ms. OpenGL ES 2 shader based drawing and conversion requires 7.82ms but the time for upload is very high and requires 18.03 which sums up to a total of 25.85ms.   It can be seen that ARM-based conversion is the fastest and sums up to 10.69ms for conversion and drawing. The CPU-based conversion requires almost double the time and sums up to 14.78ms. OpenGL ES 2 shader based drawing and conversion requires 7.82ms but the time for upload is very high and requires 18.03 which sums up to a total of 25.85ms.
  + 
  + From a performance perspective the best means for data conversion and drawing is ARM conversion with "normal" drawing. But with QML 1.2 only OpenGL-based drawing is supported. Thus, if QML is used, the best method is to do conversion and drawing with OpenGL ES 2 shaders. Finally, it should be noted that all methods are fine for live video preview because with a frame rate of 30fps the available frame time is 33.33ms.
    
  == When should multi-threading be used? ==   == When should multi-threading be used? ==




Meh.at: Meh.at - /* Discussion about hardware acceleration */
2012-05-18T19:38:07Z
Meh.at - - Discussion about hardware acceleration

			
			
			
			
		
		← Older revision
		Revision as of 19:38, 18 May 2012
		
Line 598:
Line 598:
    
  The next table gives a comparison of processing and drawing times in milliseconds (ms) for ARM, OpenGL, and [[MeeGo_Camera_VideoSurface_manipulation|CPU-based conversion]] of UYVY to RGBA data on a Nokia N9 PR 1.2 (mean over 300 runs; α=0.05 for t-test):   The next table gives a comparison of processing and drawing times in milliseconds (ms) for ARM, OpenGL, and [[MeeGo_Camera_VideoSurface_manipulation|CPU-based conversion]] of UYVY to RGBA data on a Nokia N9 PR 1.2 (mean over 300 runs; α=0.05 for t-test):
− {| class="wikitable" style="text-align: center; width: 200px" + {| class="wikitable" style="text-align: center; width: 350px" align="center"
  |-   |-
  ! scope="col" | ARM   ! scope="col" | ARM



Meh.at: Meh.at - /* Discussion about hardware acceleration */
2012-05-18T19:35:53Z
Meh.at - - Discussion about hardware acceleration

			
			
			
			
		
		← Older revision
		Revision as of 19:35, 18 May 2012
		
Line 598:
Line 598:
    
  The next table gives a comparison of processing and drawing times in milliseconds (ms) for ARM, OpenGL, and [[MeeGo_Camera_VideoSurface_manipulation|CPU-based conversion]] of UYVY to RGBA data on a Nokia N9 PR 1.2 (mean over 300 runs; α=0.05 for t-test):   The next table gives a comparison of processing and drawing times in milliseconds (ms) for ARM, OpenGL, and [[MeeGo_Camera_VideoSurface_manipulation|CPU-based conversion]] of UYVY to RGBA data on a Nokia N9 PR 1.2 (mean over 300 runs; α=0.05 for t-test):
− {| class="wikitable" style="text-align: center; width: 200px; height: 200px;" + {| class="wikitable" style="text-align: center; width: 200px"
  |-   |-
  ! scope="col" | ARM   ! scope="col" | ARM
Line 605:
Line 605:
  ! scope="col" | OpenGL + drawing   ! scope="col" | OpenGL + drawing
  ! scope="col" | OpenGL upload   ! scope="col" | OpenGL upload
  + |-
  + | 5 || 9.09 || 5.69 || 7.82 || 18.03
  |-   |-
  |}   |}
  + 
  + It can be seen that ARM-based conversion is the fastest and sums up to 10.69ms for conversion and drawing. The CPU-based conversion requires almost double the time and sums up to 14.78ms. OpenGL ES 2 shader based drawing and conversion requires 7.82ms but the time for upload is very high and requires 18.03 which sums up to a total of 25.85ms.
    
  == When should multi-threading be used? ==   == When should multi-threading be used? ==

← Older revision		Revision as of 19:43, 18 May 2012
Line 611:		Line 611:

	It can be seen that ARM-based conversion is the fastest and sums up to 10.69ms for conversion and drawing. The CPU-based conversion requires almost double the time and sums up to 14.78ms. OpenGL ES 2 shader based drawing and conversion requires 7.82ms but the time for upload is very high and requires 18.03 which sums up to a total of 25.85ms.		It can be seen that ARM-based conversion is the fastest and sums up to 10.69ms for conversion and drawing. The CPU-based conversion requires almost double the time and sums up to 14.78ms. OpenGL ES 2 shader based drawing and conversion requires 7.82ms but the time for upload is very high and requires 18.03 which sums up to a total of 25.85ms.
		+
		+	From a performance perspective the best means for data conversion and drawing is ARM conversion with "normal" drawing. But with QML 1.2 only OpenGL-based drawing is supported. Thus, if QML is used, the best method is to do conversion and drawing with OpenGL ES 2 shaders. Finally, it should be noted that all methods are fine for live video preview because with a frame rate of 30fps the available frame time is 33.33ms.

	== When should multi-threading be used? ==		== When should multi-threading be used? ==

Multi-Threaded Image Processing using Qt Camera - Revision history

Hamishwillee: Text replace - "" to ""

Hamishwillee: Text replace - "Category:MeeGo" to "Category:MeeGo Harmattan"

Hamishwillee: Hamishwillee - Fix typo

Hamishwillee: Hamishwillee - Add see also to similar articles and reference material on topic

Meh.at: Meh.at - /* When should multi-threading be used? */

Meh.at: Meh.at -

Meh.at: Meh.at - /* Discussion about hardware acceleration */

Meh.at: Meh.at - /* Discussion about hardware acceleration */

Meh.at: Meh.at - /* Discussion about hardware acceleration */

Meh.at: Meh.at - /* Discussion about hardware acceleration */

Hamishwillee: Text replace - "`" to ""`