SSTV quality

[Rob Engberts]

Hi Alex,

F.Y.I.
This morning I did a quick test with the Vizard-Meteo SSTV and compared it to MMSSTV.

The signal was noisy, but what stands out is that the MMSSTV image is sharper and therefore provided more detail. See the attached images,

Regards, Rob

[Alex VE3NEA]

Hi Rob,

It looks like MMSSTV in your experiment saw higher SNR signals than SkyRoof. When both decoders are fed with the same signal, SkyRoof retrieves more detail. Here is the latest pass of VIZARD-meteo:

[Alex VE3NEA]

Here is another A/B comparison of SSTV decoders with the same input signals:

MMSSTV:

SkyRoof:

[Rob Engberts]

Hi Alex,

I used the same input signal: MMSSTV via VB-Audio Virtual Cable.
I agree that SkyRoof comes out better in your examples ;-). However, I do get the impression that MMSSTV produces a slightly sharper image. But I’m going to run some more tests.
That said, I still think SkyRoof is an exceptionally good program!

Op dinsdag 4 augustus 2026 om 18:06:39 UTC+2 schreef Alex VE3NEA:

[Cozmo KM7DOS]

The SSTV decoder has some obvious heavy noise reduction filter on it which makes blurrier images. It would be cool if it was configurable like the external SSTV decoders are. Other than that is always looks better than the external decoders fed through VAC.

[Joseph Moore]

My SSTV images using SKYROOF only do not appear to be blurry, but maybe my vision is failing. I'm 83!!  But in this pass earlier today, I recorded what I would call 3 perfect images, and the 4th one has noise because it was near the end of the pass PLUS I have a huge tree in that direction.  Actually there are smaller trees in the direction of all three of the "perfect" images too.  I think that a good yagi tracking antenna is the most important thing.

https://www.youtube.com/watch?v=6NJH4icOkzQ

73, Dean, PU7MOJ

--
You received this message because you are subscribed to the Google Groups "SkyRoof" group.
To unsubscribe from this group and stop receiving emails from it, send an email to skyroof+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/skyroof/3559e227-4c81-4a7d-af30-8b814df6ea3fn%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Alex VE3NEA]

SkyRoof uses a so-called Wiener filter in its SSTV decoder. This filter works differently depending on the noise level. In the areas of the image where the SNR is high it does nothing and preserves all the smallest detail. In the low SNR areas where small detail is already lost, the filter uses the adaptive parameters and extracts whatever information is still present so that at least medium-level details are preserved. This is fully automatic and thus does not require a user-controlled switch.

[Robert VA3ROM]

New to the group, but long time SSTV data collector and analyser. One thing I might add that if using MMSSTV (or RX-SSTV) save images as BMP (or PNG) because JPEG uses lossy compression so the image quaility difference is even more noticeable between BMP (or PNG) because JPEG compression throws away 20-30% of the data. And lost data can't be recovered. And if you need to use JPEG, for whatever reason, use 100% quality in the MMSSTV engine settings. You'll only throw away 10% of the data, or so. The comparson images shown by Alex show heavy smoothing. I have the MMSSTV source code and am making small changes for mu own use but I do post-processing of images and comparisons with receivres and antennas using Python. Overall, for receving SSTV images, RX-SSTV is slightly superior. I use BMP for both. Post processing is either M5 (5x5 pixel window) or M9 (9x9 pixel window) depending on the determined frame SNR analysis. My phiilosophy of processing low-res SSTV images is that I'd rather have detail over data. With BMP or PNG images, I can pixel by pixel analyze the image with a Python script and make an AI fuzzy-logic processing at what the missing pixel data shoujld be.This avoids oversmoothing the image and lossy details or sharpness.

 Python has some really amazing libraries for processing images! Of course, the better the BMP image output from RX-SSTV or MMSSTV, the better the results. Of course, SSTV frames (images) with very low SNR and low correlation (quality or Q) make it more difficult to use AI fuzz-logic. We automatically use our built-in fuzzly logic when looking at an image and if there's enough data we can make an educated guess at what the image is (text is harder depending on what parts of characters are missing). These are uploaded to my my SSTV 14.230 MHz website where most of the images selected by Python are from RX-SSTV. Guy Roels, ON6MU, has done an amazing job with the MMSTV engine v1.06 to create his VB6 port of MMSSTV (Borland C++ Builder 5). I was amazed that will only some minor fixes that Embarcadero's C++ Builder 12CE can still complile Mako's, JE3HHT, 25 year old source code so it can go on and on for decades with enhancements added that standard C++ and not compiler version dependent. Not so with the YONIQ forked that used the now deprecated and horrid Builder XE (2020) and XE specific C++ routines no longer supported so it's stuck in limbo and requires a massive writting of the YONIQ addons.  

73,

Robert
FYI http://va3rom.com/SSTV/SSTV.html (RX-SSTV post-processed images) and http://va3rom.com/SSTV2/SSTV.html (MMSSTV post-processed images). 

Note: Both use SDR Console, the receiver (Airspy HF+ Discovery) and vertical antenna system fed with a 5-port active splitter) and Eugene Muzychenko's Virtual Audio Cable (VAC). VAC produces cleaner SSTV images because it uses a low‑jitter kernel‑mode audio driver with stable sample timing and bit‑accurate pass‑through which SSTV just loves. VB‑CABLE relies on a simpler user‑mode engine that can introduce jitter, drift, and resampling artifacts, which affecs SSTV decoding because it requires precise tone timing, VAC’s cleaner audio path directly results in sharper lines, fewer color errors, and fewer vertical artifacts. Now, some eyes are better than others especially if you aren't trained to know what to look for, however, mathemetical analysis of images doesn't lie. Larry Peterson, WA9TT's online SSTV comparison tool can tell you the difference mathematically:  https://petersm3.github.io/sstvcompare/ whent it's no so obvious to the eye.