KOLOR AUTOPANO PRO - AN INTERVIEW WITH ALEXANDRE JENNY
Do you have picture-stitching problems?
by Marco Trezzini
The company Kolor, one of the forerunners in picture stitching technology, was voted 'Jeune Entreprise Innovante' by the Savoy Chamber of Commerce and enjoys active support from businesses in the region to assist its development. It has also received awards from Alysé and CERA 2 (Croissance Editeur Rhône-Alpes), a software publishing cluster in France's second largest economic area. With a growth rate of +30% / per annum, its star product, Autopano Pro, is sold world-wide - 40 % in the USA, 35% in Europe and 25 % in the rest of the world. Kolor is still fully-owned by its founders; it has no debts and is proud of its excellent profit levels.
VRMAG had an interesting conversation with Alexandre Jenny, one of the founders and creators.
Meanwhile, I grab the chance to announce that in next Vrmag issue (nr. 30) there will be a review of Autopano Pro by our columnist Pat St Clair.
click here to view Kolor's team at workCan you tell us about the creators general and panoramic background?
Kolor has been created by Lionel Laissus and myself, Alexandre Jenny.
We have both worked into the video games industry before creating Kolor as engineers.
I'm a graduated engineer from a French University with a master diploma in Physics sciences.
It's been a long time I have had interest in general photography, perhaps 15 years?! (I'm 35,
so it's quite a huge percentage).
I got interested in panoramic in 1998. At that time, I moved to another town near the Alps mountains (Chambery, Savoy). During week-ends, I was used to climb the mountains. When reaching the top, you often have beautiful landscapes with a large view. So I shoot a lot... right, left... but really soon I felt a little uncomfortable because I wanted the full landscape, not only part of it. This is why I came into panoramic. There was pretty nothing out at that time or, at least, there was nothing automatic.Can you tell us more about the history of Autopano, from its initial vision, to concept, from its realization to its launch?
I was first an user of panorama stitcher. With all the mountains I climbed (and those which were climbed by friends who made shootings too for me), I quickly got a lot of stitching work. I made around 200 panoramas, each panorama taking between 1 to 4 hours to do (manual control point setting, optimization, color correction, photoshopping, etc). It was a long and not really funny job...Until the SIFT paper came out. I don't remember how I got my hand on it, but it really stroke me about the potentiality. So I wrote a SIFT control point extractor compatible with panotools product (PtGui, Hugin, PtAssembler). I found out that it worked really great. This plugin is still used a lot, it's called "autopano v1.03". The detectivity of this algorithm is incredible.
At the same time, the video game industry where I was employed got into a really huge crisis. I had to decide about my personal future. It was not really an easy step, but I believed a lot in automatic panorama creation and founded Kolor (and of course, left the game industry...).
My vision was simple: simplicity ! I had spent too much time in stitching tools and I wanted to give anyone the chance to have an accessible program. Control Points ? A photographer doesn't care about control points ! A photographer wants a stitcher that works. Exposure and color differences... I had also spent enough time into photoshopping to know that this is a major problem in panorama. Exposure changes, color shifts, etc: that was my second goal. My last vision was workflow oriented. A good stitcher needs to be efficient first. Sorting pictures out ? It seems easy to do just by looking at shooting date and regrouping images with that information. That was my vision for images groups in Autopano.How does Autopano fit into the markets dominated by Stitcher and Ptgui ?
The realization was more complicated than having the vision (It always is! Thomas Edison had said that nicely). So, first issue, the SIFT was patented. I contacted the University of British Columbia, in Vancouver to get an agreement with them. This happened pretty fast and Kolor got the SIFT license. We needed around 1 year to change completely the code behind the Autostitch tool. This was really a proof of concept tool and not at all an real software. Except for the SIFT part, quite everthing had been rewritten from scratch. The first public version of the product, the 1.0, came out the 16 december 2005. Since then, we launched 3 majors releases:
- version 1.2 : major enhancements in stability (windows only);
- version 1.3 : new rendering engine, our unique control point editor with area analysis, layer editor, history panel (mac and windows);
- version 1.4 : fisheye support with a brand new UI layout for more simplicity (windows, mac and linux)
Stitcher and PtGui are strong in the VR community. But I don't think they are so strong in the panorama community in general. In fact, that's how Autopano has made a great step for everyone: it's an easy tool. So many people that didn't think before about stitching pictures together are now doing it with our tool. We made that possible because our software is really easy and straightforward to use.
PtGui is a great tool, but it's far more an engineer tool than something for an artist (I like this stitcher a lot because I know what's behind each feature, I'm not sure it's the case for many people). I'm more reserved about Realviz Stitcher. It used to be a great tool because it was the first stitcher...But today, I really think that it missed essential points in stitching, like color correction. The central view is also not a good idea today. I found our approach of using the spherical view as central view more efficient than the QTVR view like in Realviz Stitcher.
Autopano Pro is the first stitcher to really go far beyond the panorama community. A lot of landscape photographers use our tool because it is really easy and efficient for their tasks. Architects use our tool because it's easy to straighten a building with our vertical line tool. In fact, it's only the last version of Autopano Pro that supports fisheye. It's only now that this product can be used for VR.Which is your target ?
Every photography's enthusiast. Stitching allows to enhance many part of a picture, in size (width x height ), in color depth, in field of view, in composition. It's really a new way to think a picture. My aim is to promote stitching in general, to teach people that now it works!
What are the distinctive key features ?
Efficiency, Workflow, easiness of use.
As we are based on an industrial strong SIFT detector, we have an unrivaled detectivity in panorama creation. No matter how pictures are located, no matter how shooting has been done, Autopano Pro will detect it (assuming that your images are overlapping, of course :) ). One really distinctive key feature is that our control points are ...distinctive! This comes from SIFT. Many times, I am still stunned by what Autopano Pro can see. Why did it stitch those pictures together ? Looking more closely at them ... Ho ! One is a zoom into another ! Or ho! ...It stitched this image with the other because the building in the second one could be seen reflected into the window in the first image...
Our workflow is also very well designed and makes Autopano Pro the best tool for productivity. We have customers who create 1200 panoramas in a week-end. You didn't achieve such high efficiently by chance: it was designed in the real first version.Weaknesses?
Last really cool feature is our color correction engine. It's all user oriented: you select the image that is the one to be used as color model, and others images will adjust to this one. Simple and easy to use. That's a major feature no other stitcher has.
Development has no end. As a big user of our tool, I always find some part that could be improved. We have a huge community that is always pushing us to add some new features and many of them are already planned. You can see this in the active community here.
We still have some weaknesses. Our major weakness is that we are still only a stitcher. Realviz Stitcher can do the authoring after the stitch. We cannot do that. Our HDR engine is also a bit outdated and we need to revamp it.What about Hugin and using ideas or components of Hugin ?
This is a delicate question. I really like Hugin. In fact, I talked a lot with it's primary author Pablo d'Angelo. We tried to work together to add some new common features to both products, but it didn't work the first time. We still have this project in common. About using ideas or components of hugin...There was a huge debate about that in the recent past. It was about the color correction engine created by Pablo d'Angelo, released in open source for Hugin. One month after, this color engine was integrated into PtGui, closed source software. I should not have to comment that, but I didn't find that really cool (in the first release with this engine, Pablo wasn't even quoted in the acknowledgements...).
When I started Autopano, I already had this bad feeling about a software PtGui (which means PanoTools GUI) that is using an open source panorama engine, the panotool library. It didn't take long before the first issue arose, when PtGui integrated its own rendering engine and optimizer. These parts should have been in the library and not in the software. Since then, PtGui is just using the panotools library and I do not think the authors give anything back to the open source panotools library. Because of this confusion (at least for me... I don't know if it strikes everyone as strong as it stroke me), Autopano Pro was totally designed from scratch. I could have gained 6 to 9 months development by reusing some part of Panotools, like for example the projection routines. But we did not, we made the math and solved the equations again. There is no line in common, nor even the software architecture design is the same.To conclude about using ideas or component of Hugin in Autopano: only if it were developed together. I'm totally confident about that because we found a common ground with Pablo on one project, so that's something that could happen again.Future developments?
Yes, a lot of them. HDR version 2. We need to change that. In fact, after having checked quite every software available in the HDR area, we found that no one really responds clearly to 2 problems : ghost and user oriented tone mapper. We will provide solutions to that in Autopano Pro.
We are also making a lot of developments to another branch of the Autopano brand: Autopano Server. This software is a totally automatic stitcher for online stitching. Many partners have already signed to provide stitching services to their website. This product raises a lot of new questioning about stitching in general. It also forces us to put our engine to the extreme, by forcing us to improve ever and ever the detection quality to reach the ideal of 100% good stitches whatever we throw in!
The purpose of this banner is to raise funds for a new VR community project VRMag will launch in a few months.