I’ve got a cheap computer microscope that produces some quite interesting pictures, however, its weak point is the quality of the image sensor - it’s essentially just a webcam with some additional optics.
The big problem is that the image suffers a sort of speckly noise - as webcams are prone to do under challenging lighting conditions - it appears quite random from frame to frame though, so I’m pretty sure that an average of several successive images (of a non-moving subject) would tend to let the real image come through the noise.
So what (ideally free) software can I use to produce an image where the RGB or HSV of each pixel is based on the average of the corresponding pixel in a whole set of images?
ImageJ can do this, including the suggested median filter. I have used the earlier versions of this program (NIH Image on the Macintosh and Scion Image a Windows port), to do a variety of image arithmetic tasks.
I tried Registax, but the interface is very weird and I couldn’t get it to process more than a 512 pixel square of the image - it also wants me to pick the registration point for each frame, which is unnecessary and inaccurate, because they’re all shot in the same register
I appreciate the suggestion for imagemagick, but I just don’t want to do image processing on the command line.
I’ve installed ImageJ and imported the stack of images, but the mean, median etc filters only want to work on groups of adjacent pixels per image, not corresponding pixels in the stack - they seem to be smoothing functions, not multi-image processing ones.
I found a bit of freeware called Merge that seems to do this, but only two images at a time - so I’d have to do them in pairs, then merge the pair results, then merge the results of that etc until I arrive at a single image. It’s a horribly buggy bit of software though.
The microscope is an IntelPlay QX3 - an awful little toy in all honesty, and the software interface just hurts to use, but it is capable of producing some interesting images.
Even with only a single image, many image processing programs include a “despeckle” feature which can (to some extent) remove single-pixel noise. You might need to use that if some of the noise speckles are fixed to a specific pixel.
But if they’re truly random, then a median filter is what you want. You can probably get good results with as few as three images. Basically, what you’re doing is, for each pixel, you’re taking the one from the image which is in the middle, if you put them all in order, and throwing the other copies out entirely. So if your noise really is random, then you’ll almost always be throwing away noise, and keeping a good value, so the noise won’t be contaminating your final image at all.
It’s a 60x magnified view of a human headlouse (kids. don’t ask). I thought I’d better mention the subject matter in case anyone wants to avoid seeing it.
On the left, a typical frame from the microscope, showing quite a bit of speckling - on the right, the median of 9 such frames. The improvement in quality is quite striking, I think.
The noise doesn’t entirely mask any pixel - it tends to just introduce random speckly colour tints - for that reason, I think it’s probably necessary to use at least half a dozen frames, to ensure a fairly good hit rate on the true colour value.
I have the same model QX3 (Digital Blue bought up the Intel Play line). They even have a QX5 model with a 640x480 resolution.
There used to be a website where someone had given some great tips on getting better images. I can’t seem to find it now, though. Using small supplementary bright white LEDs for direct illumination and dremeling off the lower plastic shroud around the lower body to let more light in the lens helped tremendously.
And turning off sound when using the software helps a bit too:D
The lighting definitely is one of the most significant weak points - especially on 200x magnification, where it’s completely inadequate.
I’ve created an extra user profile on my XP box just for applications (like the QX3 software) that insist on resizing the screen - I just couldn’t stand it reshuffling all my desktop stuff every time I use it.
