VideoHelp Forum
+ Reply to Thread
Results 1 to 3 of 3
Thread
  1. Hello,

    On early video camcorders a fairly brutal edge sharpening was done to compensate for the softness of the CCDs of the time. I'm referring to the early 1980's consumer cameras.

    This resulted in thick black lines running primarily vertically along edges where the dark to light transitions were above a certain threshold.

    If I was trying to recreate that effect using a convolution I might try a 1 x 3 sharpening filter (high pass) turned way up, but I would imagine this was implemented at the time using a row-based analog filter.

    Is there anyone here who may have designed or worked with that hardware at the time and who would know the exact specs that were used for that sharpening operation? Size of the window, was it single scan line based, what was the multiplier, etc.? Anything that would allow me to recreate the effect using an image processing convolution. If I can create it, I can probably undo it to some degree.

    A 1 x 3 median filter with a bit of an offset towards a maximum value helps somewhat but causes anomalies elsewhere in the image, but my goal is to create before and after image pairs (with and without the sharpened edges) to train an AI model to remove these black lines with minimal distortion.

    One solution that is unfortunately beyond my abilities would be to take a signal out of the video signal in the camera before the sharpening filter hits it and another parallel one after the filter is applied as that would create a constant stream of before and after image pairs which would be ideal for training. Otherwise, I'm hopeful that I can take a fairly high res image, as would be recorded with a modern camera, and process it using a digital sharpening filter for an output that closely matches the original analog sharpening done on early consumer camcorders and use that approach to create the before and after image pairs.

    So, sort of a two parter question, I guess: 1) does anyone who worked with or designed that early hardware know the parameters of the analog edge sharpening filters that were used that created the black vertical edge artifacts in old video footage shot with a camcorder and 2) does anyone know of an early camcorder that has an alternate video output (component, probably) from both before and after the sharpening filter hits the video signal so both signals could be digitized separately, but at the same time?

    A third question would be, does anyone here know how to engineer that, i.e. modify a camcorder to provide those outputs?

    I know it's a "Peabody, start the Wayback Machine" type of question, but I have a ton of old footage that is quite distorted by this type of edge enhancing that used to be done and I'd like to see what AI can bring to the table to reduce those edges and recreate estimations of the original edges.

    Thanks for any input,

    Ian J.
    Quote Quote  
  2. Analog video had no memory of what came before so it was done on each scanline with a simple analog sharpen filter. The amount varied from deck to deck. An easy way to simulate this in software is to downscale horizontally (to about the horizontal resolution of VHS), sharpen horizontally, then upscale horizontally. In AviSynth something like:

    Code:
    # typical 720x480/576 source
    Spline36Resize(240,480).Sharpen(1.0, 0.0).Spline36Resize(720,480)
    before/after:
    Image
    [Attachment 61963 - Click to enlarge]


    Note that Sharpen(1.0, 0.0) is essentially a 1x3 sharpen. You can use a 1D unsharpmask algorithm with the correct radius to do this without resizing.

    https://en.wikipedia.org/wiki/Unsharp_masking

    To undo it you want to reverse that process. So downscale, blur, and upscale. Of course, you don't get back the sharpness of the video before it was recorded to VHS, just a reduction of the halo.

    There are several halo reduction filters for AviSynth.
    Last edited by jagabo; 22nd Nov 2021 at 21:47.
    Quote Quote  
  3. Vertical lines can be outcome of the CCD technology - at some point they can be easily entered into saturation and as such black line is visible in output (this behavior is normal).
    Analog video processing circuits use some edge improving techniques - for example CTI for chroma improvements and for luminance similar techniques but also something called luma peaking - later similar methods was implemented in digital decoders.

    Search in google for PAL CTI and you should get some interesting results.
    You can try to do some more advanced processing - propagate signal trough edge detector, use result as mask and apply some kernel high pass filter to edge with controlled intensity - this way you should be able to avoid increasing noise and still able to simulate some analog contour improvements.
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!