Steve on Image Processing
September 28th, 2007
Major image processing product updates
At the International Conference on Image Processing last week, almost everyone who visited the MathWorks booth was already a MATLAB user. Mostly, they wanted to know "What's new?"
That question is a bit harder to answer than it used to be, because for the past three and a half years we've been releasing product line updates every March and September. Some users have the latest release, and some are still using versions that are several years old. So when someone asked me "What's new?", I usually started my response with "Do you know what version you are using now?"
The R2007b release includes a major update (version 6.0) of the Image Processing Toolbox. Here are some of the enhancements:
But that's just the set of toolbox enhancements in the last six months. How do you know what's new if your version is older than that? Go to the Image Processing Toolbox Release Notes. Click on the "Details" links under the "New Features and Changes" for the previous several releases to get a better idea of what has changed over time.
R2007b also includes a major update (version 3.0) of the Image Acquisition Toolbox. This release adds a new interactive desktop tool for configuring hardware, previewing an acquisition, and acquiring image or video data. The release also adds support for many National Instruments imaging devices.
9:23 am |
Posted in Uncategorized |
Permalink |
You can follow any responses to this entry through the RSS 2.0 feed.
You can skip to the end and leave a response. Pinging is currently not allowed.
Leave a Reply
|
|
Hi Steve — Does the added HDR support imply “raw” camera format support such as Nikon’s .NEF, Canon’s .CRW, etc? And, how about the handling of color space?
Thanks, Dave
Dave—No, HDR support means we can read a particular high dynamic range format, containing RGBe values that usually represent scene illumination in radiance units. I’m not sure I understand your question about color space.
Hi Steve — if Matlab was handling raw camera formats, I was curious to if color profile support would be similar to the TIF color space handling. I see that the image processing toolbox already has a number of color space conversion utilities available which will be useful in the HDR workflow. After all, the high-definition imagery is really nice in our Matlab figure windows, but JPG’s are still currently the web standard. ;)
Thanks!
Just to expand a bit on Steve’s comments, the .hdr format we can now read contain 32-bit floating point data in each RGB color plane, while the “raw” formats that Dave mentioned contain 12-14 bit integers in each channel. While the cameras that create those raw files are usually really great — one might say “high def” — they don’t have the same dynamic range as real HDR images. So it’s “high def” versus “high dynamic range.”
In fact, the HDR images contain such a wide distribution of illumination values in a linear light space that you need to run a “tone mapping” operation on them to visualize the data. (You can read a bit more about this on my web site: http://jeffmatherphotography.com/dispatches/2007/09/ask-dr-colors-assistant-tone-mapping-in-matlab/)
Hey Steve,
I’m enjoying the mmfileinfo and mmreader functions’ ability to work with more formats. However, they don’t report much on files with VBR or VFR. I do some work frame-by-frame that goes best if I know how many frames in advance. Any suggestions?
My first attempt was to write a binary search algorithm that pokes at the video file with im = read(readobj,framenumber) in a try / catch setup. It works, but the read times for larger files are looooong. For instance, a video with 5719 frames took about 10 minutes to locate. I can post the code if you want, but it just seems like there should be a better method all together.
Thanks for any help here,
Rob
Rob—I chatted with the developer responsible for this feature. Other customers have reported similar issues. It turns out that the Windows API we are using cannot tell us the number of frames in a variable-frame-rate (VFR) file. It will only tell us if the frame we just decoded is the last one. I expect that in R2008a there will be a slightly better way to determine the total number of frames in a VFR file, with further improvements in subsequent releases.
Hi steve, I am a student in Peking Univ. I am fond of the color space conversions in image processing toolbox, but Matlab documentation do not give the converting algorithms. Can you help to tell me how can I get the algorithms of these converting functions?
I’ll really appreciate it if you can reply to my Email. Thank you professor!
Leo—Look in the toolbox M-files for the code that does the conversion.