<div dir="ltr">Hi all,<div><br></div><div>This is a very interesting discussion and I wanted to add my 5 cents to it. John's arguments seem extremely compelling for a static view of the data. On the other hand if any processing may have to happen to the original data prior to visualization then using pyramidal TIFF may become more interesting. We are developing BisQue which serves life-sciences image data from multiple formats. For example JPEG2000, pyramidal tiled OME-TIFF or any OpenSlide supported slide data. In our case we typically work with multi-dimensional (say 4D with multiple channels) data almost never in 8 bit format and basically never stored with a lossy compression. In this case in order to render those in a browser we need to process 16 bit 5 channel data into an 8bit RGBA jpeg image. In this case lazy operation and heavy caching seems to be better than pre-processing everything, especially with many many possible versions of outputs, say different channel fusions.</div><div><div class="gmail_extra"><br><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">Serving a deepzoom client from a pyramidal tiff on a server is very<br>
slow and is rather difficult to implement. Serving a deepzoom client<br>
from a huge directory tree is hard to manage on Windows and is rather<br>
inefficient with storage.<br></blockquote><div><br></div><div>I disagree that reading tiles from a pyramidal tiff is very slow, it's is obviously slower, but with required processing it's not much slower. We can serve a 512x512 tile out form an original pyramidal TIFF image with all the processing done in about 200ms (order of magnitude slower) the first time it is requested. Our images are very large (100GB-TB), with many directories inside to represent T and Z. We cache them on the server and client sides. Second time though, the tile will be served form the server in the same 10-30ms.</div><div><br></div>In my opinion Mathieu's proposal is very reasonable in a specific use case and John's is great for a different one.</div><div class="gmail_quote"><br></div><div class="gmail_quote">Dmitry<br><div>-- <br></div></div><div class="gmail_signature">__________________________________<br><br>Dmitry Fedorov Levit <<a href="mailto:dima@dimin.net" target="_blank">dima@dimin.net</a>><br>Web: <a href="http://www.dimin.net/" target="_blank">http://www.dimin.net/</a><br>__________________________________<br><br>Center for Bio-Image Informatics:<br> <<a href="http://www.bioimage.ucsb.edu/" target="_blank">http://www.bioimage.ucsb.edu/</a>><br><br>Vision Research Lab, Electrical and Computer Engineering:<br> <<a href="http://vision.ece.ucsb.edu/" target="_blank">http://vision.ece.ucsb.edu/</a>><br><br>University of California, Santa Barbara<br>_________________________________</div>
</div></div></div>