Bart_van_der_Wolf

I'd agree with that. However, Topaz Studio is not intended as a Photoshop replacement, nor an Affinity Photo replacement for that matter. ;) It is a host program for the previous generation Topaz plugins, and for their new plugins. Some of that plugin functionality is rather difficult to achieve with PS or AP, so it's a good thing that AP allows to access that power as well. Cheers, Bart

Yes, I second that. Depending on image/resource content, one may want to aim for as lossless as possible (4:4:4) to pretty lossy (e.g 4:2:0) but with much better compression. The subsampling can help to reduce bandwidth when transferring image data, or speed-up the loading of web-pages. A means to compare the lossy effects (visually) would be very useful. Cheers, Bart

Hi, There is nothing special needed to correctly open Canon CR2 Raw files. This looks like a bug (maybe a corrupted CR2, because otherwise many Canon users would have reported it). I suggest reporting it in the Bugs forum section. Cheers, Bart

Well, there are multiple formulas in use to calculate the R/G/B weights. I don't know which one AP uses. One of them is : Y = 0.2126 R + 0.7152 G + 0.0722 B So you could use those (21% Red, 72% Green, 7% Blue) as a starting point for the R/G/B ratios in the Black&White Adjustment tool, and save that as a Preset. It's not perfect (it doesn't adjust for gamma) but it does go in the right direction. I'd have to do a better analysis of the B&W tool to find better settings. Using that preset makes it fast enough to set a personal starting point with one mouse click. Do note that the ratios used like this are dependent on which color profile is used for the document when editing. Cheers, Bart

Hi, A Black&White adjustment is not necessarily the same as Luminance, in fact, it rarely is, as you've found out. A B&W adjustment uses a certain proportional mix of RGB and CMY channels, presumably in the document's gamma (e.g. gamma 1/2.2). Luminance can be approximated by using a specific mix of R/G/B in linear gamma space. So it depends on how Luminance is calculated from the R/G/B source data, and if you can approximate that with the B&W adjustment dialog (in isolation or together with a Gamma adjustment). Cheers, Bart

Hi, I second that request, it can be a real time saver. However, it's not that simple to do correctly. Just squaring up 4 corner points will distort the original's true dimensions, because there is not enough input about the original projection distortion. Panorama stitching programs can do that correctly, but they need additional input about the focal length (e.g. from the EXIF meta-data) that was used to capture the image. Then, on the uncropped full-size image, it can try to rotate the projected image, and shift it off center (since the optical axis was not perpendicular to the subject plane), move the virtual camera position, and then crop the boundaries. So absolutely doable, not easy to do well, but it would be a very welcome addition. If done properly, it would be another improvement over what some other photo-editors produce. Cheers, Bart

Hi, Lanczos just has different trade-offs than Catmull-Rom or Mitchel-Netravali filters (or those using other Keys family filter parameters). None of them are 'proper', they just have different trade-offs. However, the ringing of Lanczos 3 (which BTW would be virtually gone with Lanczos 2, if that were to be available), can be mitigated by using a slight pre-blur (with e.g. a Gaussian blur radius of 0.25 to 0.30 times the downsampling factor) before downsampling. A down-sampling by e.g. a factor of 5, would benefit from a pre-blur of 5 x 0.25 = radius 1.25, or slightly more lossy 5 x 0.3 = 1.5. This type of pre-blur is a bit suboptimal compared to what could be achieved with much more elaborate algorithms, e.g. including the use of gamma, but it is very effective if image compression is important. It will not lose much significant detail, but it will improve the achievable lossless compression because it also suppresses the aliasing artifacts that are common with downsampling. Cheers, Bart

Hi, I second that request. Lossless file size optimization (e.g. by multi-pass compression) is very useful for lowering the bandwidth requirements of Websites and application resources. It's better to spend extra time compressing an image once, than losing time and bandwidth each time after that when an image is viewed. One could even consider optional lossy compression, e.g. 4:4:4 vs 4:2:2 vs 4:1:1 or other chroma-subsampling schemes. Stripping of meta-data is a separate issue, and should be handled with care, because it can e.g. affect/alter how images are viewed in (non-)color managed applications, and on Wide gamut displays. It will require some additional 'Expert' or 'Advanced' options in the export dialogs, and knowledge on the part of the users, but IMHO it's worthwhile (because I now have to do it with a separate application for each final image). Cheers, Bart

Hi, No, AFAIK you can't change the color/brightness of the marching ants, if that's what you mean. If you first make a rough selection, you can 'Refine' the selection with a choice of previews, e.g. an overlay in Red. Alternatively, you can make a temporary duplicate layer (in Windows: CTRL J) and either invert it or change its brightness/color, make your selection there and then, with the selection still active, select the source layer for doing the actual operation on. Cheers, Bart

Hi Lee, Yes, that's what I figured, since there are no specific 360 toggles for treating both left and right edge as part of the same image region. The only workaround (for the moment) would be to temporarily add a copy of one of the edge regions to the other side, i.e. extending the horizontal field of view beyond 360 degrees (may need to go as far as adding half an image or 180 degrees). Then taking a crop from the center (width equal to 2x height) should have smooth transitions. But it remains a bit of a kludge (and needs checking of Zenith and Nadir for visible changes). It would be useful if e.g. the Tonemapping persona would have a 360-degree option/toggle, but I understand that it would take some serious code changes. Cheers, Bart

Hi, Yes, that's correct, but I think that the OP's worry is that internally there is a potential gamut limiting operation like e.g. in DxO Optics Pro, which internally restricts gamut to Adobe RGB primaries, even when converted to a larger gamut colorspace like ProPhoto RGB (with lots of empty gamut space). It would help if one of the Moderators could confirm if anything like that, which I do not think it does, happens in AP/AD. Otherwise, we'd have to test with a very saturated image in a larger gamut output space, and compare (e.g. with ColorThink or similar 3D gamut visualization tools) with various known large gamut spaces to see if clipping/compression (depending on rendering intent) occurs. Cheers, Bart

Hi Keith, I assume that the seam in the 360-degree VR pano is where the left/right boundary of the equirectangular source image meets? Since the Tonemapping persona applies potentially strong exposure, contrast, and brightness adjustments to the image, it would have to have a 360-degree toggle to make sure that the left/right edges of the image get the same adjustments. That would certainly be a useful feature addition, something to request in that forum. The only way to avoid such seams with the current toolset, is to either do the tonemapping on the tiles before stitching, or as a workaround produce an equirectangular+ image, which has an additional image column (a duplicate of the left or right edge added to the other edge). That would produce an image that has a width of more than twice the heigth (i.e. more than 360 degrees x 180 degrees), so after tonemapping, you'd have to crop its width. The additional replicated column at the left or right can be a part of the stitching (just duplicate and rename one column of images and add them to the stitching operation) or a temporary copy/paste from the existing EXR equirectangular image before Tonemapping. The latter is probably faster since it avoids having to create control points in the stitcher (although PTGUI would have no problems with that since the images are then identical duplicates). Cheers, Bart

Hi, In addition to the current options for Lanczos 3 (separable/non-separable) which usually are fine for downsampling there are two possible improvements for upsampling or transformation/rotation/distortion possible. To avoid the risk of the gibbs phenomenon (edge ripple or ringing artifacts) a Lanczos 2 filter option should not be too hard to add, it's just a parameter for the already available Lanczos code. It would most likely preserve detail better than Bicubic, and have less risk of generating artifacts than Lanczos 3. Very useful for distortion and rotation or noisy/structured surfaces as well. Another very decent generic up- and downsampling filter would be a Mitchell Netravali filtered resampling, for which the GPU optimized code is readily available (e.g. in Nvidia's CUDA Parallel Computing Platform). Cheers, Bart

Hi Lee, I've not searched all the forum entries yet, maybe it's mentioned somewhere else? But how does the updating process go? I've noticed that the original Lensfun list is kind of dated, but that there is an updated(?) listing available on another website (maintained by Torsten Bronger). Is there a procedure that users of Affinity Photo can perform themselves, or does one have to wait for an update from you guys? Thanks for a great product. Cheers, Bart