This chapter is not intended to be a test to determine whether Zerene Software is better than Helicon Focus (it would be like trying to prove that ravioli is better than cappelletti …) or vice versa, but considerations on some differences that I have observed playing with both stacking platforms.
To perform this analysis, I used as subjects the tiny scales of a beautiful butterfly Papilio ulysse acquired with a 50x Mitutoyo lens, Pseudobrookite crystals, with a 5x Mitutoyo lens, and a bee, with a 2.5x Mitutoyo QV lens.
Butterfly scales: the stack is composed of 330 images, with an acquisition step between one frame and the next set to 0.5µm; the illumination used was based on continuous light, through two 20W LED panels and exposure time set to 1″. The cylindrical diffuser is “home-made” with white drawing paper.
Pseudobrookite mineral: the stack consists of 286 images, with an acquisition step between one frame and the next set to 10µm; the lighting used is based on continuous LED light, using the illuminator OGGLAB DB120EB and exposure time set to 1/60 “.
Bee: the stack consists of 259 images, with an acquisition step between one frame and the next set to 20µm, the lighting used is also based on the use of OGGLAB DB120EB illuminator and exposure time set to 1/60 “.
I used the SONY A7 RII camera, 200mm tube-lens with Raynox150, and Ultra RAIL mini v2. The computer used to perform the calculations is a Laptop ThinkPad Intel® Core ™ i7-8850H CPU @ 2.60Ghz with 32Gb of RAM and Windows 10 Pro 64bit.
The algorithm used to compile the final images is based on PMax (in Zerene) and Method C (in Helicon), algorithms that are quite similar:
PMax is a “pyramid” method. It is very good at finding and preserving detail even in low contrast or slightly blurred areas. It’s also very good at handling overlapping structures like mats of hair and crisscrossing bristles, nicely avoiding the loss-of-detail halos typical of other stacking programs. But PMax tends to increase noise and contrast, it can alter colors somewhat, and it’s liable to produce fuzzy “inversion halos” around strongly contrasting objects. As a parameter selected in the Preferences menu, I leave the “Grit suppression” option selected by default, it makes a small sacrifice in saving fine detail, in exchange for the benefit of getting significantly less noise in the final image [from: HERE ].
Method C uses pyramid approach to image processing dividing image signals into high and low frequencies. Gives good results in complex cases (intersecting objects, deep stacks), though increases contrast and glare [from: HERE ].
The photos are captured in RAW mode (standard .ARW format in SONY camera) at 16bit, converted to .jpg (8bit) and .tif (16bit). The stacked image is not modified, not filtered to improve the contrast, the sharpening, and no noise removal filter was applied.
For each calculation, the time required to obtain the stacked image was determined.
For the butterfly scales dataset, a first image test was calculated using HF by importing the ARW format directly into the software. This program allows you to read the native RAW formats of the main photographic cameras, convert them temporarily to .TIF format, and save them in .DNG format while keeping intact the colorimetric profile present in the initial RAW. It is important, therefore, to consider having the necessary space to save the temporary files created on your hard disk. The calculation to obtain the final image, using all 330 frames took 15′ 21.07″.
To compare the performances and the results of both software, the RAW files are converted to .JPG and .TIF formats. ZS does not provide RAW converters, then third-party software must be downloaded separately from the net, or acquire specific software for image processing and conversion like, for example, Adobe Lightroom or CaptureOne, or use the software provided by the camera manufacturer. Since 16-bit TIFF format files retain the image metadata and have more bits per pixel than those captured by the majority of today’s cameras, this process preserves all the image quality inherent in the RAW format.
The time required to calculate the final stacked image can be seen below (these values can vary in function of the processor and the graphic card installed on your computer):
JPG (8bit) Butterfly Scales (330 images)
TIF (16bit) Butterfly scales (330 images)
TIFF (16bit) Pseudobrookite (286 images)
TIFF (16bit) Ape (29 images)
It is impressive to observe the difference in calculation times between the two programs. Helicon Focus turns out to be monstrously fast in providing the final output: there is no possibility of comparison between the two platforms.
This is for what concerns the quantitative performances. And what about the quality of the result?
As specified before, no editing tools were used to modify/improve the stacked image, even though both software have particularly efficient and advanced editing tools.
Let’s go to explore the distribution of the pixels for the three bands (red, green, blue) composing the 16bit image of the butterfly scales.
The image exported from ZS produces an image with a distribution of pixel values that tend to be slightly more shifted toward light tones than HF. This is due to the possibility to export the image with the Retain extended dynamic range option selected. The latter (HF) returns a darker image, more saturated in the blue band, as indicated also by the lower mean and median values.
The histogram on the left is calculated on the image obtained from ZS, the one on the right with HF. The blue band has a higher saturation at the extreme values in HF, and this affects the relatively higher brightness in the image visible on the right.
Pushing the magnification of both images to 300%, we can observe a distributed noise that permeates the image created with ZS, while the image obtained by HF is more homogeneous, smoothed, and without (or less) noise. Apart from the presence of noise in ZS, there are no differences in detail in both images.
Overlaying the ZS and HF images and creating a GIF animation, one can highlight the differences in geometry between the two outputs. Evidently, the algorithms used by the two platforms, although similar, perform a slightly different reconstruction of the final stacked image.
The image of the Pseudobrookite crystal also lends itself to interesting comparisons as we are in the presence of linear parallel structures and flat reflecting surfaces that can highlight eventual problems of calculation and algorithm interpolation.
Once again, we can observe a diffused noise in the ZS output image, noise that is absent in the HF image. The absence of granulation allows to better highlight the finest details presents on the flat surfaces of the crystals, as well as the eventual presence of impurities (dust, inclusions…) on the reflective surfaces. In addition, the shadow present on the crystal surface appears localized in two different positions, more advanced in ZS and backward in HF.
A reflective surface creates a kind of glaring on a crystal surface on the HF image, which is not present in the ZS output.
At the level of color distribution (unchecking the Retain extended dynamic range option when saving the .TIF format in ZS), the statistical values of the mean and median are very close, observing in the tails of the histogram that the HF output is slightly more shifted towards the light tones, and towards the dark tones in ZS.
The last subject analyzed is the bee. With the Retain extended dynamic range option selected in ZS output (then making an automatic level adjustment that reduces the contrast and increases the brightness of the image), we obtain a stacked image that is lighter in color (left) than the image obtained with HF (right).
Calculating the respective histograms, we have confirmation of what we observed visually. ZS output shows a histogram that presents tails in the red, green, and blue channels, while the result obtained with HF output has colors that are more shifted in the left sector of the color space, creating a darker and contrasted image.
As visualized in the previous images of the butterfly scales and crystals, the output image of the bee calculated by ZS, shows a diffuse visible graininess compared to the image calculated by HF.
The ZS image, due to the relatively large noise, loses some details, especially in correspondence of homogeneous dark surfaces. While lighter surfaces show slightly higher levels of detail than HF.
HF output presents a «crispier» image losing locally slightly in detail but increasing the contrast in the presence of linear structures as, for example, in the bee’s bristles. This contrast in some cases is visible locally as a sort of jaggedness, which is less visible in ZS (structures that are eventually noticeable zooming at 300%, but invisible at full resolution).
As already said in the introduction, this short and simple test does not want to put one software in front of the other since they both have features that make them both interesting.
Last but not least, both programs have the possibility to visualize and export 3D images for more or less volumetric visualizations. This option has a very relative interest, and the software is certainly not purchased to access this feature since there are much more advanced programs to perform three-dimensional rendering (such as Pixel4D for example). In any case, the viewers allow you to appreciate the subject in oblique view and export in animated GIF, .mp4, or stereo-pair format.
Helicon Focus from the point of view of the calculation performances is unbeatable, and the time “saved” allows eventually to correct the (rare) errors present in the image. Moreover, the user interface is extremely friendly, allowing you to work and get results immediately, without having any knowledge of the program itself. The possibility of being able to work directly with files in RAW format avoids an intermediate step, accessing the result more easily.
Zerene Stacker is much slower but gives access to parameters allowing more personalized/fine-tuned results. This requires spending a bit of time studying the platform and the options (but the documentation is very well done and complete), even if the look and feel is a bit “outdated”. Personally, I find the output image concerned by important noise also if not really affecting the result (but I may have omitted to check some parameters in the preferences of the software to remove this noise).
As far as prices are concerned, HF turns out to be less expensive (but not too much) to have access to the Pro license than the Prosumer Edition of ZS; both once purchased the license are lifetime, without having to pay additional prices during the upgrades.
You can find the prices by clicking on the following links:
Trial versions are available (30 days) allowing access to all the functionalities and having a global idea of the software.
Below you can find some links to sites where are discussed the differences between the two software, the pros, and cons. Everyone will be able to choose the platform that best suits your need, functions also of the available personal budget.
(Some) useful links:
the first three links are part of THE reference in extreme macro techniques. A lot of videos and information (and fun). Don’t miss his tutorials!