Messages

751

Feature Requests / Weighted Average Texture Blending Mode

« on: July 25, 2012, 02:26:16 PM »

Using average blending mode, often the edges of photos can be apparent in the final texture if there were changes in lighting.

Would it be possible to implement a 'weighted' average blend mode, where the weight is 1 for non masked areas, 0 for masked areas and the 'area' outside the image, and a user defined gradient ,i.e. number of pixels, between the two. The gradient should begin at the edge of the mask/image so that masked areas do not begin to re-appear.

I think this should at least soften the egdes? Or any other algorithm to achieve the same result!

I tried doing this myself by blurring my masks in photoshop, but these are always converted to black/white on import.

Mosaic blend seems to do a better job of blending joins between photos, but because I have such changes in lighting (I didn't take the images!) average mode is better at creating a uniform 'light distribution'.

Currently using 0.90 pre-release 1534 pro 64 bit.

752

General / Re: Ornate Ceiling

« on: July 20, 2012, 04:49:43 PM »

I am awaiting permission to release the data and will let you know.

In the mean time, I have come up with a solution although not yet attempted it:

I will divide the photo set into small chunks that align correctly with respect to each other, even though there may only be 3 or 4 images per chunk. I should then only need to specify 3 markers per chunk to bring them all into the correct local coordinate system.

I think it's going to be a long slow process...

753

General / Re: Ornate Ceiling

« on: July 20, 2012, 01:14:42 PM »

In the event that there is no solution using the tools currently available in photoscan, could I propose a new tool?

"Estimate initial camera positions using ground control points"

I think that 4 appropriately distributed GCPs per photo should be sufficient assuming internal camera parameters are known, and even these could be estimated with more control points.

754

General / Ornate Ceiling

« on: July 20, 2012, 12:14:24 PM »

I am trying to align photos of an old ornate ceiling, which although is very colourful and detailed, is also repetitively patterned, so photos are mostly misaligned.

I have a wealth of 3D information as the ceiling was also surveyed with a laser scanner, and the photos were taken in a grid so i know roughly what the camera positions were, and the orientation is generally the same (although not always).

The ceiling is 8m x 8m, and I have ~150 images in about 12 rows of 12 photos.

Will it help the alignment if I set accurate 'ground' control markers using vertices from the laser scan point cloud, and how many points per photo would I need to specify?

Will it help to provide approximate camera positions, bearing in mind they will be very approximate?

Using the ground control image pair preselection method seems like an obvious choice, and this does rely on having camera positions. Does this also require camera orientation, and does it consider the marker positions?

I have tried adding markers (maybe not enough), camera positions (maybe too approximate), and orientations (maybe just wrong!) and not yet had much luck.

Does anyone have any suggestions on workflow regarding the above options? Or maybe breaking it down into chunks i.e. one chunk per photo strip.

Attached images (originals are 6MP) give an idea of what i'm up against!

I do not need to generate geometry, just align photos at this stage.

Thanks for any advice anyone can offer!

755

General / Re: Help with Texture Map and Zbrush

« on: July 20, 2012, 11:41:01 AM »

Never used zbrush, so can't comment on that, but once you have cleaned your mesh you can import back to photoscan and generate texture on the cleaned mesh within photoscan.

756

General / Re: too much processor consumption?

« on: June 16, 2012, 11:28:23 AM »

I'm using win7 not linux, and not asking for help, just reporting!

757

General / Re: too much processor consumption?

« on: June 15, 2012, 12:48:16 PM »

I managed to recreate this problem too (not deliberately!)

I have a basic scene - 26 photos 14MP each and have not yet even generated geometry.

Before aligning photos I went through and set 23 markers but did not give them coordinates in GC. Then I performed align photos.

It took a long time to display the points (22k points) after aligning and now just moving the mouse over the 3d viewport sends processor usage to 100% for about 5 seconds.

I haven't ever used the approach of settings markers before alignment, and on this occasion the alignment is completely screwed up anyway. Until now I had been aligning the photos and then creating and coordinating the markers with no issues.

PC is i7 2700k 16GB GTX 580

PSZ file too large to attach but have extracted the doc.xml file

758

Bug Reports / Re: OBJ File Standard

« on: May 04, 2012, 12:53:37 PM »

Sorry just rechecked, and didn't need to add anything to import to blender with texture/material already attached.

3dreshaper is very powerful, great for pointcloud to mesh, but still has a lot of bugs. I haven't tried geomagic, i think it's probably better but possibly beyond our budget at the moment.

759

Bug Reports / Re: OBJ File Standard

« on: May 03, 2012, 12:13:00 PM »

I am using 1423.

To import to 3D Reshaper I needed to add both lines to import the mesh at all.

To import to Meshlab I didn't need to add any lines and the mesh was correctly textured.

To import to Blender I needed to add the usemtl line for the texure to be applied on import.

Once imported to 3D Reshaper, the UV mapping appeared incorrect, but since it is fine in photoscan, blender and meshlab it is probably their problem?

760

General / Re: aerial photography?

« on: May 02, 2012, 07:27:57 PM »

If you own a good HD-Cam: Just take a walk around an object and cut the recorded video into roundabout 100 (sequenced) pictures. It works - not really good but it works if you want to scan a high number of objects in short time. Take a look at another software (not released yet) which shows this technique: http://www.vi3dim.com/ . For sure PhotoScan will give you better results if you use 600 (!) HD-cam-pix

This can work quite well as a way to quickly generate alot of images. I use ffmpeg http://ffmpeg.org and the following command to split into still images:

Code: [Select]

ffmpeg -i input.avi -r 1 -s 1920x1080 -f image2 frame_%04d.jpg
Thought I'd share as it took me a while to figure it out.

input.avi =: movie file - doesn't need to be avi though
-r 1 = rate i.e. 1 image per second
-s 1920x1080 = dimensions of output images

761

Bug Reports / OBJ File Standard

« on: May 02, 2012, 07:19:43 PM »

OBJ files exported from Photoscan do not import into 3D Reshaper (http://www.3dreshaper.com/).

I contacted Technodigit to see if this was a bug with their software, with an example OBJ, but they state that the OBJ is not compliant and should contain the following lines:

Code: [Select]

...
vt 0.454957 0.497448
vt 0.507275 0.474474

---------Code to add------------
g Mesh1
usemtl Textured
------- End code to add------------

f 1/10 2/11 3/12
f 1/10 8/17 6/15
f 1/10 6/15 5/14
...http://www.technodigit.com/forum/viewtopic.php?f=6&t=20

Is this a shortcoming of Photoscan, or is the OBJ standard just not well defined?

James

762

Feature Requests / Multiple photo view for setting markers

« on: January 19, 2012, 07:26:17 PM »

It already looks like photoscan should support a multiple document style interface, but i can't see how to get two photos open alongside each other in order to set control point markers easily on photo pairs.

Also when alternating between photos, the zoom level in one photo is affected by the zoom level in the other photo.

If photos could be opened alongside each other, with independent zoom levels, then it could be much faster to specify markers.

763

General / Re: How to spped up process?

« on: January 19, 2012, 06:52:20 PM »

Something I do to speed up the process is use all photos for alignment, but then disable maybe 50-75% of them before building geometry.

My 'logic' is that the overlap is only needed to calculate camera position and scene structure. Once that is calculated you do not need multiple photos on the same thing in order to build the gemoetry.

However I never worked on such large data sets, and am not familiar with all the algorithms involved so this may not be an appropriate workflow for some applications!

764

General / Re: Is photoscan professional right for me?

« on: November 22, 2011, 01:58:12 PM »

Does anybody have an example of an terrestrial-based orthophoto (ideally building interior or exterior) created in photoscan professional that they would be so kind to share?

It's the main thing we are likely to use if we buy the professional edition, but it's the one thing we can't evaluate in the trial version!

Or can anyone suggest a more appropriate software package?

Thank you

765

General / Re: Is photoscan professional right for me?

« on: November 11, 2011, 02:24:54 PM »

There is the difference between texture creation and orthophoto generation. Creating orthophoto allows to use several additional options, e.g. hole filling. Also the photos for orthophoto generation are selected by special algorithm improving the quality of resulting photomap. And also the size of the orthophoto could be greater than one of the texture.

This sounds great, but is there any way to evaluate it before spending the money?! Maybe by seeing a couple of example orthoimages created by photoscan of something like a building facade - maybe from the building in the "Building Modelling" tutorial? Or the ability in the trial version to export orthophoto with a watermark, or time limited functionality etc?

Thanks