Popular Content

Showing content with the highest reputation on 01/18/2018 in all areas

  1. 1 point
  2. 1 point
  3. 1 point
    This is a plugin intended to help remove the backgrounds from images. I wasn't sure where to put it, but decided to make it an Adjustment. The controls are: Match Color is normally the color to be erased, but it can optionally be the color to keep. It can be either the User Match Color, or the Primary or Secondary Color. User Match Color is the Match Color when User Match Color is specified. Hue Tolerance determines how closely the pixel's hue must match the Match Color's hue. Saturation Tolerance determines how closely the pixel's saturation must match the Match Color's saturation. Value Tolerance determines how closely the pixel's value must match the Match Color's value. RGB Tolerance specifies the maximum allowed Euclidean distance between the pixel's RGB color and the Match Color's RGB color. Gray Upper Limit (S×V) specifies the threshold for product of the saturation and value below which the Match Color or pixel's color will be classified as gray. Gray does not match any hue except gray, unless the Gray Matches All Hues option is selected. Portion of Non-Erased Color to Preserve determines how much of the "non-erased" color should remain in the pixel. The color and alpha of the pixel are adjusted to account for the degree to which the pixel color matches the Match Color. Specifically (assuming the Match Color is being erased), the color is adjusted so that the alpha is as small as possible while keeping the same color if the pixel is alpha-blended into a Match Color background layer. This can be used to achieve a softer edge. Though its not really intended to be used when the Erase Non-Matching Pixels option is selected, it does work. The erased pixels will all be the Match Color, with appropriate alphas. The value can range from 0, to entirely erase the matching pixel, to 1, to preserve as much of the pixel's color as possible. Normally, it will probably be either 0 or 1. Gray Matches All Hues specifies that colors classified as gray match all hues. Erase Non-Matching Pixels specifies that matching pixels are kept and non-matching pixels are erased. Here is the source: Here is the icon: Here is the plugin: HsvEraser.zip I will happily consider any suggestions to make the interface clearer. I will also appreciate any spelling corrections. EDIT: Fixed several bugs. Changed some of the wording in the interface. Added the RGB Tolerance. Changed the version to 1.1.
  4. 1 point
    With the help of MJW's advice, I came up with an amazingly good result, which I thought I'd share. What I realised was that the noise (animated snow) is almost entirely different on each image. That is, snow particles rarely overlap between images. So taking the difference between two otherwise identical images results in just the snow particles on a black background. I reasoned that this could be subtracted back from the combination of the two original images, factoring in the fact that each contributes half of the intensity. So, if the two background images are BG1 and BG2: Create two copies of each, as alternating layers. Set the top layer to Difference mode and merge down to give BG1-BG2. Set the top layer of the lower pair to 127 Opacity (50%) and merge down to give an average of the two, BG1+BG2. Now set the topmost of the two combined layers to Difference AND 127 Opacity and merge down, which I'll call BG1/BG2. The result is a superbly clean and sharp rendition of the background. Frankly, I was astonished at the quality of the result. Repeating the same process with the foreground images, I can Difference the FG and BG results to get just the desired overlay image. Analysing, step 2 creates an image of the noise (snow particles) in both layers. Step 3 gives an average of the two layers, which also halves the intensity of any snow particles that don't appear in both images. Step 4 subtracts half of the snow particles from the averaged image, effectively removing them. Of course, if there are any snow flakes that appear in the same location on both images, they won't be removed, but this can be handled with additional source images. Unfortunately, I have other images where the noise is more fog-like, and I suspect this process won't work. In which case it will be back to simple averaging. Nonetheless, this is exactly the kind of layer algebra that I was looking for. Yay!
  5. 1 point
    You went through all of that and then painted it white?
  6. 1 point
    Very Nice TR! That's a way I haven't seen before to re-color things. If I do blend modes I normally have my color above the object and set the color to a blending mode. I liked how you did yours the other way around though. I personally think using selection tools to make a cut-out is the fast way, so it is good to point out that that is one way of doing things. However, using selection tools often doesn't work for pictures that have more difficult colors or patterns to try and select, and it sometimes leads to uglier cut-outs vs other methods. So I still recommend cutting things out by duplicating the layer and erasing with the line/curve tool or by the normal tracing an object to make a cut out. Either way, it's nice to see the variety of answers here.
  7. 1 point
    Here's a technique I developed to choose colors for the outside of our home. Use Magic wand and other selection tools to select the clothes to be recolored copy clothes to a new top layer. make top layer B&W use Curves in adjustment menu to change contrast towards white but make it look natural select empty part of top layer and invert selection - (don't forget to pick replace or you'll look like an idiot if you make a video) click on image (original) layer and delete clothes by cutting or pressing delete change top layer blend mode to multiply!!!! create a new layer bottom layer and use the Paint Bucket to paint it the new color This technique works with clothed people, naked people, planes, trains, automobiles, and the occasional exterior of a house. BTW - we painted the house white. image from wikimedia public domain license
  8. 1 point
    That's a really beautiful sig, DS. Dreamy even.