eronald Posted March 9, 2009 Share #81 Posted March 9, 2009 Advertisement (gone after registration) Trust me on this one. I've worked in high-end digital imaging (not Photoshop) for over 15 years. Good code can go a long way to retain image quality, but once you perform the kind of transformation as in the example you gave, none of the pixels in that image are the same, as when they were captured. Trust needs to be earned. I have been mystified by maths since I went to school. Never understood any of it, although I have mostly made my peace with the need for it. Haven't noticed getting better at it with time - why should I assume *you* have got better because you worked so long at it ? I haven't got any smarter with time. Ok, so you're not a rubber fetishist like I am. So, let's try another experiment. Take a 1Kx1K image. Oh, so small. We need to print larger. Our image needs more pixels. Let's Interpolate it up to 100Kx100K . Now do the barrel correction. If I have a couple of percent distorsion, and I scale up the grid by a factor of a hundred because I need 100x dots to print,how much information will I destroy while applying the transform? Edmund Link to post Share on other sites More sharing options...
Advertisement Posted March 9, 2009 Posted March 9, 2009 Hi eronald, Take a look here S2 and my impressions. I'm sure you'll find what you were looking for!
wildlightphoto Posted March 9, 2009 Share #82 Posted March 9, 2009 Trust needs to be earned. I have been mystified by maths since I went to school. Never understood any of it, although I have mostly made my peace with the need for it. Haven't noticed getting better at it with time - why should I assume *you* have got better because you worked so long at it ? I haven't got any smarter with time. Ok, so you're not a rubber fetishist like I am. So, let's try another experiment. Take a 1Kx1K image. Oh, so small. We need to print larger. Our image needs more pixels. Let's Interpolate it up to 100Kx100K . Now do the barrel correction. If I have a couple of percent distorsion, and I scale up the grid by a factor of a hundred because I need 100x dots to print,how much information will I destroy while applying the transform? Edmund Edmund, I don't understand what you're trying to say here. Are you asking a question or making a point? Consider how small and inexpensive a fisheye lens is compared with something such as the forthcoming 17mm TSE. For Canon it is about $650 vs $2500 and 12 onces vs. 29 ounces. It occurs to me that if a fisheye lens could be made that is very high resolution and the sensor is also high resolution, one could afford to lose some of the sensor resolution when correcting for the distortion of convergence in the fisheye photo. A perfectly valid trade-off. Consider the other implications of this trade-off, for example the sensor's noise properties that accompanies the very high resolution. This scenario also assumes a lower margin for exceptional circumstances, such as when the client is so delighted with your work he wants a wall-sized print for the corporate reception area. Link to post Share on other sites More sharing options...
AlanG Posted March 9, 2009 Share #83 Posted March 9, 2009 A perfectly valid trade-off. Consider the other implications of this trade-off, for example the sensor's noise properties that accompanies the very high resolution. This scenario also assumes a lower margin for exceptional circumstances, such as when the client is so delighted with your work he wants a wall-sized print for the corporate reception area. I get your point but it seems that if we take a format such as 36x48 or the new slightly larger sensors, there is so much resolution on the sensor that this might be a reasonable approach for extreme wide angle. Especially considering the lack of rectilinear extreme w/a lenses for this format. In any case, they don't have to be complete fisheye lenses. I'm just thinking that if a lot of barrel distortion can be left in the formula, there might be other advantages in the lens... such as it will offer a solution that is not available any other way. I doubt if the Canon 15mm 2.8 fisheye is the end all be all of what is possible, yet it is a very good lens considering its size, weight and price. Make a better one for a larger format and it might surprise us. Remember, anamorphic lenses are used for several wide screen movie systems. In any case, at one time I shot all of my architectural images on 4x5 because architects, builders, and interior designers put a premium on large prints for their offices. But that changed with the economics of shooting FF 35mm digital. And the fact that large prints still look quite good from this format. The S2 and MF cameras may produce images that are "better" when blown up to 24x36 and beyond, but is there enough additional income from clients who will want these large prints, and can they appreciate the quality enough to justify the expense, lack of flexibility, and the slower usage of such cameras? At one time I would have said yes. But not today. On the other hand, my fine art work is made up of hi res stitched photos producing large files with the hope that I can make high end large print sales. Link to post Share on other sites More sharing options...
AlanG Posted March 9, 2009 Share #84 Posted March 9, 2009 How many typical lenses challenge the resolution of the sensor? I suspect that you will see more resolution loss with a better-than-typical lens. Still might be better than the typical lens, but you will see a loss of image detail. Forget the lens. And we can step away form being theoretical. One can take any given image, apply a bit of distortion, and judge for oneself how much detail is lost. If it is acceptable, then a lens design with that much distortion in the opposite direction would also be acceptable to you. As you have just proven that it can be corrected and still meet your requirements. I think technology may soon offer some other solutions. A new Sony camera and a cheap General Imaging camera both offer "sweep" automatic panorama stitching. One just holds down the shutter, in the case of the Sony it then shoots a series of images as you rotate the camera as far as 224 degrees, and then it stitches them together into one image. I can see a pro version of this incorporated into a camera that is used on a pano head being incredibly handy. Especially if it has software to make 120 degree or so high pixel count rectilinear images. I don't see why someone couldn't simply develop software that allows you to create still panos from hi def movies. And surely there will be things we haven't even thought of. Link to post Share on other sites More sharing options...
stevelap Posted March 9, 2009 Share #85 Posted March 9, 2009 Don't get me wrong, the main Leica 'premier' dealers in London (Red Dot, RG Lewis, etc.) are great and I enjoy a visit to these shops, but they are a world away from the kind of London dealers (The Pro Centre, Teamwork, etc.) who have day to day dealings with the professionals that the S2 was meant to be targeted at. It seems strange to me that Leica would think that the traditional cosy dealer is the appropriate place to sell the S2. It's a bit like being expected to go to your local garden centre to buy a combine harvester. I wonder if it's a tacit admission by Leica that, contrary to the pre-launch marketing spin, they actually expect the S2 system to sell mainly to affluent amateurs. I'd agree that whatever system Leica intend to put in place for S2 support will need to stand comparison to that offered to the pro's by the likes of Pro Centre, Fixation and others (using London as an example). Calument (Drummond Road) is already a Leica dealer of course. They did say to David, Guy and others at PMA though, that sales and support for the S2 would be handled by the Leica dealer network, and the S2 minisite LEICA Camera AG - Aktuell still mentions a web portal for an "exclusive service for professionals" which "will be the home of our professional online customer service" (although they also said starting in January 2009). So perhaps they have something in mind that will satisfy the support expectations of the pro's without bypassing existing dealers completely? Link to post Share on other sites More sharing options...
thrid Posted March 9, 2009 Share #86 Posted March 9, 2009 Trust needs to be earned. I have been mystified by maths since I went to school. Never understood any of it, although I have mostly made my peace with the need for it. Haven't noticed getting better at it with time - why should I assume *you* have got better because you worked so long at it ? I haven't got any smarter with time. What do you want me to do? Publish my resume and school papers? If you go to the doctor and he tells you that you have a continence problem, you're not going to believe him because you didn't study medicine? Link to post Share on other sites More sharing options...
felipe-m Posted March 9, 2009 Share #87 Posted March 9, 2009 Advertisement (gone after registration) What do you want me to do? Publish my resume and school papers? If you go to the doctor and he tells you that you have a continence problem, you're not going to believe him because you didn't study medicine? thrid, you've made a good point. I work a bit on digital imaging and in my opinion you are entirely right. If someone with no background on the topic and general issues with math chooses to ignore or distrust your straight forward explanation, that's not your problem but his, so no worries there. (And this is not meant to be personal comment). Transformation and interpolation will go a long way, especially with modern algorithms, but the bottom line is that, no matter how good they are, the result is not what the sensor registered. It is not really a pure RAW file anymore. So a system that gives you a file that is closer to what the sensor registered by needing one step less of in-camera processing is at least that much better. This may not be critical sometimes, maybe mostly RAW purists will care much about it, but in general a "RAWer" file means more potential, flexibility and overall quality. Link to post Share on other sites More sharing options...
wildlightphoto Posted March 9, 2009 Share #88 Posted March 9, 2009 Forget the lens. And we can step away form being theoretical. One can take any given image, apply a bit of distortion, and judge for oneself how much detail is lost. If it is acceptable, then a lens design with that much distortion in the opposite direction would also be acceptable to you. As you have just proven that it can be corrected and still meet your requirements. Alan I haven't proven anything, I suggested it. Link to post Share on other sites More sharing options...
thrid Posted March 9, 2009 Share #89 Posted March 9, 2009 I don't see why someone couldn't simply develop software that allows you to create still panos from hi def movies. This already exists. Some of this software is proprietary, others commercial. Link to post Share on other sites More sharing options...
AlanG Posted March 10, 2009 Share #90 Posted March 10, 2009 Alan I haven't proven anything, I suggested it. When I said "As you have just proven," I wasn't speaking of you or anyone in particular. I simply meant that if anyone adds a bit of pincussion distortion to any existing picture (that meets one's quality standards) and does not see enough resolution stretching to be bothered by it, then that person (not necessarily you) has proven to himself that he would be happy with a lens that required the same amount of software correction in order to compensate for a lens' barrel disortion. Link to post Share on other sites More sharing options...
guruguhan Posted March 13, 2009 Share #91 Posted March 13, 2009 I have just been reading through this thread, I haven't finished it yet, but wanted to ask one question. Firstly, this is the first time I'm hearing that S lenses will work on the future R camera (great)...Does this include the leaf shutter lenses (any info on whether they will work)? If the 70mm S lens was used on an r body (with a 36x24mm sensor)...what focal length would it be? Link to post Share on other sites More sharing options...
BJNY Posted March 13, 2009 Share #92 Posted March 13, 2009 The focal length of a lens is permanent. Link to post Share on other sites More sharing options...
dfarkas Posted March 13, 2009 Share #93 Posted March 13, 2009 I have just been reading through this thread, I haven't finished it yet, but wanted to ask one question. Firstly, this is the first time I'm hearing that S lenses will work on the future R camera (great)...Does this include the leaf shutter lenses (any info on whether they will work)? If the 70mm S lens was used on an r body (with a 36x24mm sensor)...what focal length would it be? This hasn't been verified 100%, but it would be really cool to have the ability to use high-sync speed leaf shutter lenses on a 35mm DSLR. Because the R10 and the S2 will share so much technology, my guess is that the S lenses will function on the R body. But, we will have to wait and see for this one. A 70mm lens will be a 70mm on an R body. No factor. David Link to post Share on other sites More sharing options...
lct Posted March 13, 2009 Share #94 Posted March 13, 2009 ...If the 70mm S lens was used on an r body (with a 36x24mm sensor)...what focal length would it be? The focal length would remain the same but the equivalent field of view would be about 87.5mm. Link to post Share on other sites More sharing options...
carstenw Posted March 13, 2009 Share #95 Posted March 13, 2009 The focal length would remain the same but the equivalent field of view would be about 87.5mm. Huh? A 70mm on FF 35mm looks like a 70mm. It is on the S2 that it looks like something else, in 35mm terms, ie. a 56mm lens or so. Link to post Share on other sites More sharing options...
lct Posted March 13, 2009 Share #96 Posted March 13, 2009 Huh? A 70mm on FF 35mm looks like a 70mm. It is on the S2 that it looks like something else, in 35mm terms, ie. a 56mm lens or so. The diagonal of the S2's sensor is 1.25x longer than that of a FF 24x36, so S2 lenses will have a 1.25x crop factor on the R10. Then 70x1.25 = 87.5mm EFoV. Same when you use a 50mm lens on your M8: 50x1.33 = 66.5mm EFoV. The focal length remains the same but the equivalent field of view is multiplicated by the crop factor. Right? Edit: Not so sure though as the EFoV of a 70mm lens is 70/1.25= 56mm on the S2. I must be wrong somewhere, so David above and Georg below are probably right. Edit2: You also of course Link to post Share on other sites More sharing options...
georg Posted March 13, 2009 Share #97 Posted March 13, 2009 Nope: 24x36 -> M8 = 1.33x - 70mm equals 93mm 24x36 -> S2 = 0.8x - 70mm equals 56mm 24x36 -> R10 = 1,0x - 70mm equals 70mm Link to post Share on other sites More sharing options...
peterv Posted March 13, 2009 Share #98 Posted March 13, 2009 The diagonal of the S2's sensor is 1.25x longer than that of a FF 24x36, so S2 lenses will have a 1.25x crop factor on the R10. Then 70x1.25 = 87.5mm EFoV. So much confusion I think lct has the right answer here. Link to post Share on other sites More sharing options...
lct Posted March 13, 2009 Share #99 Posted March 13, 2009 ...I think lct has the right answer here. Which one? Link to post Share on other sites More sharing options...
mhoersch Posted March 13, 2009 Share #100 Posted March 13, 2009 The diagonal of the S2's sensor is 1.25x longer than that of a FF 24x36, so S2 lenses will have a 1.25x crop factor on the R10. I must be wrong somewhere, ... Right, you are wrong there. The crop factor on the S2 is 0.8, not 1.25, since the sensor is bigger than 35mm, not smaller. That's why a 70mm lens will look and feel like a 93mm on an M8 (1.33) and like a 56mm on the S2 (0.8). Link to post Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.