Canon has today announced a new ultra-high resolution CMOS image sensor that packs approximately 250 megapixels into an area smaller than a United States postage stamp.
A comparison of inexpensive, consumer level 3D cameras in terms of the specifications that matter for 3D photography and video. There are other cameras besides these but these are affordable and available from many vendors. Another day I will look in to specifications of higher end cameras, such as the Sony TD10/TD20 3D video cameras.
Update: Since posting this item it appears that most of these cameras are in the process of being discontinued and you can find some great half price deals right now. It is likely that new 3D cameras will be introduced in a few weeks at the 2013 Consumer Electronics Show.
|Fujifilm W3||Lumix 3D1||Toshiba Camileo Z100|
|Stereobase||6.5 cm||3.0 cm||3.0 cm|
|Focal length||35-105mm 35mm equivalent||25-100mm 35mm equivalent||Fixed lens, 4x digital zoom in 3D, 35mm equivalent unknown|
|Stills||2x 10.0 MP Sensor3D Resolution
7.2 MP in 16:9 ratio
8.9 MP in 3:2 ratio
10.0 MP in 4:3 ratio
Stills recorded in full size side by side MPO format
|2x 12 MP Sensor3D Resolution
6 MP @ 16:9
8 MP @ 4:3
Stills recorded as full size
|2x 5 MP sensor3D Resolution
4 MP @ 16:9
(2D 5M 2592 x 1944, and an “interpolated” 16 MP 4608×3456)
Stills recorded in full side-by-side JPEG
MP4 encoding3D HD Resolution
two separate video streams recorded as 1280×720/24p for left
1280×720/24p for rightEncoding
AVCHD or MJPEG
in either AVCHD or MP4. Sensor is progressive but video is encoded as 1080i – end result is basically the same as 30p.
AVCHD or MP43D Resolution
960 x 1080 for left
960 x 1080 for right
MP4 encoding,3D file format is one half side-by-side formatmeaning 960×1080 for each halfFeatures external mic input plug
|LCD||Glasses free 3D||2D only||Glasses free 3D|
|Image stabilization||No, CCD imager||Yes, CMOS imager||No, CMOS|
|Battery||user replaceable||user replaceable||user replaceable|
|Sony Bloggie 3D|
|Focal length||16:9 stills and video: 47mm (35mm camera equivalent)
4:3 41mm (35mm equivalent)
|Stills||2x 5.15 MP sensor
(3.1 MP @ 16:9)
(5 MP S 4:3)3D
2 MP (1920x1080x)Encoded as full size MPO side by side images
3D: 1080/30p onlyEncoding
MP4Encoded in half size side-by-side 3D format for 960 x 1080 resolution per side
|LCD||Glasses free 3D|
|Image stabilization||Yes, CMOS imager|
|Battery||internal, not replaceable|
An interesting observation – some of these cameras have 1920×1080 image sensors but actually cut the image in half when encoded into video. The reality is they are not 1920×1080 but 960 x 1080 x 2.
In addition, the frame rate offers additional temporal resolution.
Let’s compare the image quality in terms of actual resolution, as well as resolution in time by calculating a “mega pixels per minute” rate:
- Fujifilm W3: 1280 x 720 x 2 x 24 fps = 44.2 MP/minute
- Lumix 3D1: 960 x 1080 x 2 x 30 fps = 62 MP / minute
- Toshiba Z100: 960 x 1080 x 2 x 30 fps = 62 MP / minute
- Sony Bloggie 3D: 960 x 1080 x 2 x 30 fps = 62 MP / minute
- Generic 1280 x 720 x 30 fps = 55 MP / minute
- Generic 1280 x 720 x 60 fps = 110 MP / minute
Interestingly, depending on many factors, your highest image quality might come from 720 x 60 fps because it delivers more potential information to your eyes over time.
Shooting 3D with my two Kodak Playsport Zx3 cameras, I think the 720/60p dual camera view looks on par with the 1080p/30p view. But that is also because after editing and 3D processing, the output of a 1080p video often ends up as a 960x1080p side by side video (as needed, for example, to upload to Youtube).
Figuring out which is best can get complicated!
“Rolling shutter” is a video image distortion that commonly appears when using modern CMOS-based DSLRs – and also CMOS-based consumer camcorders. It manifests as a wobbly or skewed image. In the simple case, a fast pan left or right causes vertical lines to become slanted and bent. In a more typical case, portions of the image can appear wobbly if the camera – or the subject – are moving left or right.
I’ve elevated the following to its own blog post; it first appeared within the following post about the new Canon SX40IS. I also wrote about rolling shutter before.
I’ve been discovering some issues with rolling shutter on my Lumix GH-2 when using long telephoto settings that are, for me, more problematic than fast pans. Specifically, if I use the 45-200mm zoom at the 200mm setting (think 400mm full frame equivalent lens), and then switch into the ETC extended digital teleconverter mode where it isolates just 1920×1080 pixels (multiply by 2.6 times) giving a 1040mm (full frame) effective lens, the very slightest motion produces skew and wobble in the image.
Last week I shot a scene using this feature – since a 1040mm equivalent lens is compressing a huge amount of atmosphere, the thermal refraction occurring in the image made the image wobble. And sure enough, that resulted in rolling shutter issues even though the camera was locked down securely on a tripod!
Some day … an electronic global shutter will be added to CMOS sensors, I suppose. Until then, for long range video shooting, I prefer CCD imagers.
Is rolling shutter a problem? Some claim its not if you merely plan your images in advance. For those who can plan their images, this strategy may work fine.
But I shoot a lot of live events that I do not control and or which there can be only minimal planning. I am finding that DSLRs are great for shooting video:
- For wide angle views, including handheld shots
- For producing narrow depth of field
- For convenience and small size (relative to a prosumer camcorder like the XH A1 or HMC-150)
- Where you do not need real time audio monitoring and audio controls
- Where you do not need a motor controlled zoom
- For excellent low noise video images (especially at low ISOs)
For long telephoto shots, I find myself fighting rolling shutter far too often. I cannot control the wind. I cannot control the air temperature that causes thermals and refractions, making the image move around. As you can see in the linked articles, below, some people like all the wobbliness! I don’t!
Real video cameras have the following advantages:
- CCD imaging (hopefully!) and no rolling shutter
- Audio controls without adding on extras (I use a Beachtek audio mixer and external mics with my Lumix GH-2)
- Motorized zoom control
- Better auto focus. The GH-2 tends to hunt when shooting video so I end up using manual focus.
DSLRs (although not the GH-2) tend to suffer from false image artifacts created by aliasing and moire patterns in the images
On a typical modest screen HDTV (mine is 42 inches) it is very difficult to tell the difference between GH-2 video and XH A1 (once the A1’s noise issues are addressed – I’ll add that item to another post).
DSLRs are a fantastic tool for video, but they are not yet the be-all solution for video. But they are a wonderful additional tool to have for video shooting.
- iPhone video enhanced by rolling shutter effect (dpreview.com)
- IPhone’s Rolling Shutter Captures Amazing ‘Slo-Mo’ Guitar String Vibrations (wired.com)
- How rolling shutter technology can warp your photos (gadling.com)
- iPhone 4 Rolling Shutter Captures Amazing Guitar String Oscillations (VIDEO) (blippitt.com)
- Power zoom lenses for DSLR cameras (coldstreams.com)
Older video cameras used CCD-based image sensors. For various reasons, that I will explain in a moment, camera makers have largely switched to using CMOS-based image sensors on low end and even some low end semi-pro video cameras (like the Panasonic HMC-40). All of the new digital SLRs (DSLR) still cameras that also shoot video using high resolution CMOS sensors too.
But there’s a nasty problem with CMOS-based sensors that can cause the video image to resemble a shaky bowl of jello, as illustrated in this video comparison between the Canon HV20 (CMOS-based) and the Panasonic SD5 (CCD-based):
The problem is that the CMOS image sensor is read “line by line” from top to bottom. If the image changes during the read out, then one line may be slightly offset or shifted from the previous line. This is known as a “rolling shutter” and creates the peculiar “jello effect” since not all lines (or rows) in the image are lined up with all the others.
Old movie film cameras used a physical shutter that open and shut exposing an entire frame all at once. (Although, there are some issues with that too which we can ignore for now.)
CMOS reads the image row by row while CCDs read the entire image all at once and do not use a “rolling shutter” – hence, no jello on CCD cameras.
So why use CMOS instead of CCD? The basic reason is that CMOS uses less power and produces less heat and is less expensive. As image resolutions have increased, the size (and cost and power) of the CCD imager has gone up.
Camera makers, especially in the consumer market, are in an arms race to each have cameras featuring ever more pixels. This means most have switched to CMOS because, apparently, most consumers do not care about the jello or do not encounter it often enough.
In the DSLR world, most any camera with 10 or more megapixel resolution is CMOS. CMOS works great for most still photography. But when these DSLRs are used for video, they too suffer from awful “jello”.
There are other things to consider too – CCDs may show vertical streaks in photos when there are very bright lights (sun reflected on water, stage lights, etc) in the image. For higher resolutions, CCDs need to be physically larger – and more expensive.