“To visualize the vision” makes me feeling confused for a long time. Bunches of questions buzzing in my head while I started playing with the microscope.
- What am I supposed to do if the source is already viewable to the eyes?
- Would it make sense just to stretch the images, scale them, detect some edges and adjust the hues?
- What information am I supposed to digg out underneath the ready-to-see visual pieces?
- Would it be too scientific if it’s just about analyzing the visual paramaters of source images?
- Am I able to do that without acknowledgement of the scientific facts behind them?
However, apart from all above, I come up with two different approaches or “flavors” initially.
- To create a multifunctional tool which provides multi-dimensional access to the source. It could be a helper tool for the scientists or whoever wishes to digg information from the source. The end user would be able to tweak over different parameters (scale, rotation, color, brightness, contrast, and a lot more stuff we see under the Image-Adjustment menu of Adobe Photoshop) to look into the source for what they need. The visualizer must create an interface which is flexible enough to make possible all these tuning and manipulation.
- To create a specific experience which guides the viewer through specific emphasis on certain aspects of the source. By emphasis it means to focus on, give different weight to, cast strong contrast on one or more feature of the source so that in certain aspect it creates a twisted illusion of the source.
For the first one my worry is that to be “flexible” could easily become an excuse for creating a visualization without thinking of or understanding the viewer. One can just pile up the algorithms in a computer graphics textbook and throw all the stuff to the viewer, and let the viewer to explore through. This kind of “all-in-one” solution might be suitable for expert viewer with certain target in mind, yet would also get the common audience easily lost in the data. On the other hand, it is in fact taking more effort to implement all possible dimensions in tweaking, which makes it even more infavorable to me.
The problem of the second one is that it would be boring. Given the context that I’m using a low-res toy computer microscope to collect raw samples, there’re actually not much data to dig for. The interesting thing about the toy microscope is that despite of the low-res results, it does provide 3 different magnitudes (10x, 60x and 200x). Good enough to start with!
It is not to say that these two approches are mutually exclusive. Both of them require collections of raw images in different scales to form a usable dataset. Some of them could be created digitally, but now that the microscope comes with 3 different scaling settings, I could collect more precisely the raw data in the aspect of size and detail.
Another interesting fact is, with a fix-sized window, while you’re getting much more detail at the 200x level, you lose the big picture. The only chance you can see the whole sample is when you switch back to 10x or with your naked eyes. To collect data to cover both needs, I need to take picture at:
- 10x. 1 picture (full view, low detail)
- 60x, 4 pictures (medium detail)
- 200x, 16 pictures (high detail)
So it’d approximately 20 images for each sample. The visualization should be able to create a relatively high-res image from the collage of more detailed however smaller images.
As for the visualizing, I would pretty much like to combine the two approaches, which would be:
- Providing the viewer capability of examining the samples in different detail levels (zooming in/out, rotating and transforming maybe);
- Always providing scaling information while the viewer is observing the sample, letting him/her be aware of current detail level;
- Focusing on view experience rather than observation result of the viewer. Viewer behavior would be recorded as sample data to visualize too in order to reveal one’s visual interest on certain sample. These data could be revealed in a heatmap or 3D histogram over the 2D sample.
- Collecting accumulated viewer behavior data to form an aggregated result of visual interests.