Image Stablization with HV10 at DVinfo.net

William Porter · May 22nd, 2007, 01:53 PM

I have been using the HV10 now for a couple of months as a helmet cam for mountain biking. I have been doing helmet cam stuff for a few years now (SD) using a Sony DCR-PC9 that used electronic image stabilization. Now while not looking for sidewalk smooth images of mountain biking, I like the way the PC9's image stabilization took the "sharp edges" off of the motion/bumps. The HV10's OIS does not seem to do as good of a job. My gut feeling is it does not have as much corrective range or response time as the Sony making it less effective in a mountain biking environment. I suppose hand holding the camera in a boat while filming a water skier would be a similar environment. Maybe the HV10 is optimized for correcting movement handheld while standing still type movements.

Here are two samples. They were shot using the same bike with the same suspension setup. The rider was the same in both (me) with the same fatigue level. (I have found the body does the biggest part of the stabilization for MTB vids. When you start getting really fatigued you get bounced around more and your video suffers. Fatigue was not an issue in these videos, so the "body dampening" was the same.)

Here is the footage with the HV10. This is a different trail than shown with the old setup up, but this trail should be easier from a stablization standpoint - File size is 127MB

http://www.mountainbikebill.com/videos/SART07-Web.wmv

(Disregard the darkened edges, I was trying out a Tiffen protective filter that showed up on the footage but not on the LCD.)

Here is footage from the Sony PC9 setup. This trail should be harder from a stabilization standpoint, particularly from minute 2 to 3. File size is 95MB

http://www.mtbbill.com/videos/Cannell06.wmv

I would like others opinion on my assumption.

I would love to find out there is some way to change (firmware update/setting) the optimization of the OIS system, but I not holding my breath. What are some of options available to take some of the edge off in post (PPRO2)? I have tinkered with the 2D3 plugin that comes with premiere but I think I have to change the workflow as the interlacing messed with it.

-Bill Porter
www.mountainbikebill.com

Duane Steiner · May 22nd, 2007, 03:16 PM

I do in car videos with the camera hard mounted in the car. I just got a HV20 and find that the OIS is not as good as my SD cameras (Panasonic GS250 and older JVC). It is pretty much unwatchable. I hope to be able to add some dampening material to my mount or try a different one.

Found this piece of software that might help but have not tired it yet: http://www.guthspot.se/video/deshaker.htm

Austin Meyers · May 22nd, 2007, 04:20 PM

i think the biggest issue isn't so much the OIS as it is the rolling shutter issue with the cmos chip. the warping in the frame is the cmos not being able to capture it all quickly enough and that adds to the "shaking" it's especially exaggerated in 24p mode

George Ellis · May 22nd, 2007, 04:27 PM

Bill,

First, I took the WMV, upres'ed it to 720p and then downres'ed it. So, it did make it look like a Tony Hawk Lipstick with all the res changes. This was my second attempt to deshake it using Deshaker. Deshaker is a VirtualDub plugin and Harold Linke creates a VirtualDub FEX plugin for Studio and Liquid that allows VDub plugins. Long way around, but you may be able to find an equivalent with PP or at least use VirtualDub. So, the foggy results are...

http://www.versatilemediasolutions.c...MBBDeshake.wmv

This attempt, I changed to block size from Discard motion from 300 to 500 and Set the Deep Analysis to < 25%. Lots more tweaking and better done with a 1080 source. A fused the results and then fit them to a SD with a 2D effect to increase the size in the timeline to remove the black space.

Yep, not perfect, but a start to an experimental workflow...

Oh, this is 9:20 to 9:31 in the clip.

Duane Steiner · May 22nd, 2007, 07:58 PM

Here is video from inside my car. I am going to try and add some rubber or foam to see if I can isolate the camera. I think that part of the problem might be HD in general. With so much more resolution, everything is magnified. Also might be the rolling shutter effect. What does everyone think?

http://www.socalspeedzone.com/incartest/incartest.html

William Porter · May 22nd, 2007, 10:34 PM

Quote:

Originally Posted by Duane Steiner

Here is video from inside my car. I am going to try and add some rubber or foam to see if I can isolate the camera. I think that part of the problem might be HD in general. With so much more resolution, everything is magnified. Also might be the rolling shutter effect. What does everyone think?

http://www.socalspeedzone.com/incartest/incartest.html

I certainly hope it is not a HDV thing. Just for reference, here is an in-car vid from a friend of mine. I believe it was done with a PC9 as well. You can definitely see that EIS is working well. The car does not seem to much smoother than yours but there is certainly a huge difference in stablization.

http://www.mtbvideos.net/video/lagunatraffic.mpg
(Video by Pete Fagerlin)

-Bill Porter

Duane Steiner · May 22nd, 2007, 11:14 PM

For comparison these videos were taking in car with the same mount from my SD cameras: http://www.socalspeedzone.com/pages/...107/index.html

George Ellis · May 23rd, 2007, 03:58 AM

Actually, it may be a HV10 thing. Douglas Spotted Eagle posted about problems with the HV10 and extreme sports POV. Skydivers were getting unusable footage because of what appears to be element shake. So far, I have not seen anyone with a HV20 have the exact same experience, but they have not been pushed as hard. It looked like Duane's in-car, only worse. Bill seems to have a camera with less issues though. Maybe the vibration frequency is different enough?

Mikko Lopponen · May 23rd, 2007, 05:27 AM

You need a camera with no rolling shutter for high speed action. It has nothing to do with hdv.

It also seems that the hv10 has some problems with its ois implementation?

William Porter · May 23rd, 2007, 10:23 AM

Quote:

Originally Posted by Mikko Lopponen

You need a camera with no rolling shutter for high speed action. It has nothing to do with hdv.

It also seems that the hv10 has some problems with its ois implementation?

As I understand OIS, you need a bit of "margin" (sorry for lack of a technical word) around the sensor so that when the image is stablized you don't get moving black bands on the edges of the footage. Is this correct? If so, could it be that with the push to get to HD in these cameras, there is very little margin on the sensors so the stablization can only correct so far before the black edges would appear.

Since the HV10 can shoot in SD as well, I should have a bunch of "margin" for the OIS to use. This weekend I'm gong to try and find a bit trail and make several runs on it recording in both HDV and SD modes to see if there is any noticeable difference between the two in the same camera. I'm I way off the mark here?

-Bill Porter
MountainBikeBill.com

Mikko Lopponen · May 23rd, 2007, 10:28 AM

Quote:

Originally Posted by William Porter

As I understand OIS, you need a bit of "margin" (sorry for lack of a technical word) around the sensor so that when the image is stablized you don't get moving black bands on the edges of the footage. Is this correct? If so, could it be that with the push to get to HD in these cameras, there is very little margin on the sensors so the stablization can only correct so far before the black edges would appear.

Nope. EIS uses margins over the picture so it can scan around. OIS doesn't, that's why its called optical.

Duane Steiner · May 23rd, 2007, 10:35 AM

Quote:

Originally Posted by William Porter

Since the HV10 can shoot in SD as well, I should have a bunch of "margin" for the OIS to use. This weekend I'm gong to try and find a bit trail and make several runs on it recording in both HDV and SD modes to see if there is any noticeable difference between the two in the same camera. I'm I way off the mark here?
MountainBikeBill.com

I am going to do the same with my HV20 in the car.

May 22nd, 2007, 01:53 PM	#1
William Porter Tourist Join Date: Mar 2007 Location: Vista, CA Posts: 3	Image Stablization with HV10 I have been using the HV10 now for a couple of months as a helmet cam for mountain biking. I have been doing helmet cam stuff for a few years now (SD) using a Sony DCR-PC9 that used electronic image stabilization. Now while not looking for sidewalk smooth images of mountain biking, I like the way the PC9's image stabilization took the "sharp edges" off of the motion/bumps. The HV10's OIS does not seem to do as good of a job. My gut feeling is it does not have as much corrective range or response time as the Sony making it less effective in a mountain biking environment. I suppose hand holding the camera in a boat while filming a water skier would be a similar environment. Maybe the HV10 is optimized for correcting movement handheld while standing still type movements. Here are two samples. They were shot using the same bike with the same suspension setup. The rider was the same in both (me) with the same fatigue level. (I have found the body does the biggest part of the stabilization for MTB vids. When you start getting really fatigued you get bounced around more and your video suffers. Fatigue was not an issue in these videos, so the "body dampening" was the same.) Here is the footage with the HV10. This is a different trail than shown with the old setup up, but this trail should be easier from a stablization standpoint - File size is 127MB http://www.mountainbikebill.com/videos/SART07-Web.wmv (Disregard the darkened edges, I was trying out a Tiffen protective filter that showed up on the footage but not on the LCD.) Here is footage from the Sony PC9 setup. This trail should be harder from a stabilization standpoint, particularly from minute 2 to 3. File size is 95MB http://www.mtbbill.com/videos/Cannell06.wmv I would like others opinion on my assumption. I would love to find out there is some way to change (firmware update/setting) the optimization of the OIS system, but I not holding my breath. What are some of options available to take some of the edge off in post (PPRO2)? I have tinkered with the 2D3 plugin that comes with premiere but I think I have to change the workflow as the interlacing messed with it. -Bill Porter www.mountainbikebill.com

May 22nd, 2007, 04:20 PM	#3
Austin Meyers Contributor Join Date: Aug 2005 Location: austin, tx Posts: 300	i think the biggest issue isn't so much the OIS as it is the rolling shutter issue with the cmos chip. the warping in the frame is the cmos not being able to capture it all quickly enough and that adds to the "shaking" it's especially exaggerated in 24p mode __________________ Compositing is my life right now. IMDB page: http://www.imdb.com/name/nm3596782/

May 22nd, 2007, 04:27 PM	#4
George Ellis Trustee Join Date: Feb 2004 Location: Suwanee, GA Posts: 1,241	Bill, First, I took the WMV, upres'ed it to 720p and then downres'ed it. So, it did make it look like a Tony Hawk Lipstick with all the res changes. This was my second attempt to deshake it using Deshaker. Deshaker is a VirtualDub plugin and Harold Linke creates a VirtualDub FEX plugin for Studio and Liquid that allows VDub plugins. Long way around, but you may be able to find an equivalent with PP or at least use VirtualDub. So, the foggy results are... http://www.versatilemediasolutions.c...MBBDeshake.wmv This attempt, I changed to block size from Discard motion from 300 to 500 and Set the Deep Analysis to < 25%. Lots more tweaking and better done with a 1080 source. A fused the results and then fit them to a SD with a 2D effect to increase the size in the timeline to remove the black space. Yep, not perfect, but a start to an experimental workflow... Oh, this is 9:20 to 9:31 in the clip. Last edited by George Ellis; May 22nd, 2007 at 04:29 PM. Reason: include time

May 22nd, 2007, 03:16 PM	#2
Duane Steiner Regular Crew Join Date: Feb 2006 Location: CA Posts: 195	I do in car videos with the camera hard mounted in the car. I just got a HV20 and find that the OIS is not as good as my SD cameras (Panasonic GS250 and older JVC). It is pretty much unwatchable. I hope to be able to add some dampening material to my mount or try a different one. Found this piece of software that might help but have not tired it yet: http://www.guthspot.se/video/deshaker.htm

May 22nd, 2007, 07:58 PM	#5
Duane Steiner Regular Crew Join Date: Feb 2006 Location: CA Posts: 195	Here is video from inside my car. I am going to try and add some rubber or foam to see if I can isolate the camera. I think that part of the problem might be HD in general. With so much more resolution, everything is magnified. Also might be the rolling shutter effect. What does everyone think? http://www.socalspeedzone.com/incartest/incartest.html

May 22nd, 2007, 11:14 PM	#7
Duane Steiner Regular Crew Join Date: Feb 2006 Location: CA Posts: 195	For comparison these videos were taking in car with the same mount from my SD cameras: http://www.socalspeedzone.com/pages/...107/index.html

May 23rd, 2007, 03:58 AM	#8
George Ellis Trustee Join Date: Feb 2004 Location: Suwanee, GA Posts: 1,241	Actually, it may be a HV10 thing. Douglas Spotted Eagle posted about problems with the HV10 and extreme sports POV. Skydivers were getting unusable footage because of what appears to be element shake. So far, I have not seen anyone with a HV20 have the exact same experience, but they have not been pushed as hard. It looked like Duane's in-car, only worse. Bill seems to have a camera with less issues though. Maybe the vibration frequency is different enough?

May 23rd, 2007, 05:27 AM	#9
Mikko Lopponen Major Player Join Date: Mar 2006 Location: Helsinki, Finland Posts: 393	You need a camera with no rolling shutter for high speed action. It has nothing to do with hdv. It also seems that the hv10 has some problems with its ois implementation?