Although great care was taken not to compress images at any point in the test sequence, minimal-loss compression had to be used to prepare the pictures for posting. In most cases, this should not interfere with your ability to evaluate the results. If you want to see the original image set, with NO compression,
download the complete image set as a ZIP file. This file is about 4.5 megabytes in size, so I doubt that any but the hardcore SSTV hardware and software types will want to take the time. Remember, you MUST view the pictures that follow in UNCOMPRESSED form or you cannot evaluate the results.
This shows the reference image set with no added noise. Here the original source image [S] can be compared with the image display provided by the Pasokon [P], Chroma Pix [C], and WinPix [W] systems. Note that all three systems did an excellent job with the image under noise-free conditions. However, if you look VERY closely, some differences can be seen. The sharpest reproduction was provided by the ChromaPix system, followed by WinPix. By comparison, the Pasokon image is just a bit "soft". This differential in image sharpness is constant at all noise levels and will not be noted again.
However, both the ChromaPix and WinPix versions reproduce the picture just slightly darker than the original (look critically at the lower background areas) while, in contrast, the Pasokon is virtually identical to the original in terms of brightness. As we shall see later, this slight darkening appears to be an artifact of the DSP filtering in both the ChromaPix and Winpix software and is characteristic of the reproduction noted at all noise levels. The green and black bar at the base of the block is an artifact of the BMP to JPEG conversion and should be ignored. As expected, with a "clean" signal, all systems successfully auto-started without any problem. At this noise level, any differences are subtle and, if you can declare a "winner", it is subjective and the margin is small.
This minimal amount of noise did not inhibit auto-starting with any of the systems and the only feature contributed by the noise is a slight granularity in the facial tones and, less obviously, in the background. This effect is subdued a bit by the relative "softness" of the Pasokon picture [P2] and is emphasized by the comparative sharpness of the ChromaPix rendition [C2].
Although noise is easily heard at this level, all three systems autostarted without difficulty. Noise is evidenced by much greater granularity throughout each image. A VERY careful examination will show that the filtering provided by the ChromaPix software [C3] is superior to the other two systems but that the differences are obscured by the progressive "softness" of the WinPix [W3] and Pasokon [P3] pictures.
At this noise level, the ChromaPix [C4] and WinPix [W4] images auto-started cleanly but the Pasokon image [P4] had to be line-started. ChromaPix provides the best noise filtering, followed by WinPix and then Pasokon. In the case of the WinPix image, some very slight color aliasing (blue-green pixels) can be seen in the forehead area and on the cheek on the right side of the image.
At this noise level none of the images triggered on the VIS code and, had the mode not been preset correctly, the pictures would have been lost. All, however, did trigger successfully on the sync pulses and the complete image was displayed in each case. The previous ranking of filter effectiveness (ChromaPix, followed by WinPix, followed by Pasokon) is even more evident in this case. The blue-green color aliasing that was just starting in the forehead and cheek areas with WinPix in the previous set of pictures is now even more evident and very slight traces of the same effect can be seen in the ChromaPix and Winpix images (look carefully at the forehead area above the nose). At this noise level, the voice signal could probably not be copied.
This set demonstrates the most frustrating situation when chasing DX with marginal signal levels. You can clearly hear the SSTV signal down in the noise, but can you get the picture to display? WinPix was the only system to copy the entire frame. The Pasokon system was able to randomly trigger on a few lines but nothing useful was obtained. As usual, ChromaPix did the best filtering job, but could not display the entire image.
Since the goal of these tests was to look at image display in the presence of noise, both the ChromaPix and WinPix programs were run with the maximal DSP filtering available. This image set is included to compare display with these two systems, with and without the DSP filters engaged. The results are quite surprising. The images represent the Level 5 noise test and the unfiltered examples are on the top row with the filtered examples below
.The WinPix image set (right) shows what you might expect - a significant noise reduction when the filter is operating. If you really need to dig a signal out of the noise with WinPix, using the eight-pole filter has some obvious benefits, although it does introduce that blue-green aliasing previously noted on the forehead and cheeks.
By comparison, the ChromaPix filter (left image set) provides NO obvious benefit and I feel it actually degraded image quality just a bit. My suggestion would be to not use the filter function when using this software!
Early on, a reference was made to the fact that both ChromaPix and WinPix displayed the image just a bit darker than the source image. If you compare the top and bottom image sets, you can see that this is primarily a result of the DSP filtering of the signal.
All three systems managed to display full pictures in all but the most severe noise situation. There are certainly observable differences, particularly at the highest noise levels, but if you expected spectacular differences in performance, you were probably surprised at the results. All three systems obviously deliver good value and, for most operators, decisions as to which system to use will probably be governed by other program features.
I intend to keep the test image tape and, sometime soon, will test external hardware and DSP filters. The basic test methodology worked so well, I also plan a test series to look at how the different modes function in the presence of noise.