Photiosity - Manual (Version 1.5)
Photiosity is an application to sharpen blurry and shaky photos
Rolf Geprägs Software Development, Hamburg
MacOSX 10.7 (Lion), 64 bit.
Photiosity uses the OpenCL library. OpenCL provides a good optimization for the CPU (SSE, multi-core) and also for the graphics card (GPU). Photiosity runs on Macs with and without an OpenCL graphics card (eg: HD Graphics 3000), but is significantly faster on Macs with an actual GPU.
Photiosity uses a computationally intensive method to deblur the image. It is possible to achieve results that are far better than the image sharpening in many image editing programs. However, this requires the interactive cooperation of the user.
Now Photiosity uses an optimized algorithm, which is much faster than the previous one (1.0.3). Nevertheless the old algorithm is included and can be easily reactivated (switch off the Fast Calculation Method checkbox). Photiosity uses the CPU as standard computing device.
The Divergence is automatically determined by Photiosity during the sharpening process (Auto in Settings), but is sometimes not exactly predictable. Therefore, you can affect the quality of the result (disable Auto-checkbox) by manually changing the divergence value.
If the difference between the original and the sharpened image is hardly visible, then the divergence value is too low (eg 0). If the sharpened image has high contrast areas (in the example: the white areas, which are to large), the divergence (example: 10000) is too high:
The theoretical sharpness of the calculated image is shown as a thin black curve in the graphics on the page Basic Settings. This value is approximately a function of the sum of differences in luminance of neighbouring pixels in the image. The visual sharpness of the image is a very subjective term, which isn’t necessarily proportional to the theoretical sharpness. Nevertheless the theoretical sharpness gives a strong hint on the visual sharpness.
The term divergence is a kind of measure that indicates the difference of the sharpened image from the original image. Therefore a small divergence value (-> 0) means that the sharpened image approximates the original one. A high divergence value (-> 10000) escalates the sharpening process. The result of these two extreme cases is often unsatisfactory. In the first case, it gives you an image that differs hardly from the original source. In the second case, it gives you a sharpened image which is characterized by high contrast and distortions.
How so often, the best divergence value lies in the middle (4000 - 7000).
The mentioned graphics shows the current value of divergence as a blue line. The red straight line indicates the given divergence. During the sharpening, the blue line is approaching the value of the red line, and so the current divergence value converges to the given divergence value. It is important to know, that the sharpening process is only active, if the current divergence is not higher than the given divergence (blue line <= red line).
When the Auto-checkbox is set, Photiosity tries to calculate the optimal value for the divergence, during the sharpenig process. The value can be changed at any time (even during the calculation). This requires only to switch off Auto.
PSF is an abbreviation of “Point Spread Function”. Principially, it is a visualization of the blurring that leads to the unsharpness in the image. It shows how the picture of one pixel would look like, after applying the assigned blurring. Photiosity tries to calculate the PSF in order to determine the original sharp image from the PSF and the blurred image. For this calculation an highly optimized variant of the iterative “Lucy Richardson Method” is used. Therefore the PSF of Photiosity is not comparable to the PSFs of other blind deconvolution programs.