I've already opened a thread about pixel shift motion detection methods, but this time I have a different idea for monochrome only images. The current way of masking motion effect in pixel shift images would work perfectly is pixels were entirely free of noise, including free of quantization noise, but unfortunately there is random noise (thermal noise, photon noise, read noise etc). For example, when I decrease motion detection threshold in RT, it confuses some pixel noise with motion, and if I increase the motion detection threshold it leaves out some of the pixel values that changed due to motion in the image. The reason is that pixel values contain random noise from exposure to exposure, and it is not possible to distinguish between pixel noise and image motion, at least it's not possible without using AI. But I don't like the idea of detecting motion from differences between pixels. You dont get artifacts with Rawtherapee, except in rare cases you have to tweak the settings.I know. You dont get artifacts with Rawtherapee, except in rare cases you have to tweak the settings. Would that method mitigate what would otherwise be color motion artifacts from the classic pixel shift operation?Are you suggesting skipping motion correction? You can't get around motion correction without accepting ghost images or blurryness. Stack the 4 monochrome images to reduce noise. create another monochrome image - blue channel only create another monochrome image - red channel only ![]() create a monochrome image from the 4 exposures using the green channel only When converting a pixel shift image to monochrome, the color image is created first, and then it is converted to grey scale, color artifacts show in the gray scale image.īut what if pixel shift frames would be handled differently for monochrome images? False colors are created because of lost color consistency between frames when color pixels are create from 4 exposures. Pixel shift is known for turning image motion into color artifacts.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |