7

Like in any other kind of signal processing, the relationship is Nyquist's theorem. An image is a discrete sequence of samples of a continuous signal. If the original signal has frequency components higher than half the sampling rate, then there will be aliasing. To put that another way, if you look at the real-world size of a pixel, any details smaller than ...


6

If I get correctly what you are asking you basically just need to find the G in this equation: $$Image_{out} = Image_{in}^G$$ This could be easily solved as $$G = \frac{\log{Image_{out}}}{\log{Image_{in}}}$$ Because usually gamma is applied in a uniform fashion on the image, you can just pick any two non zero pixel values (one for source and one for ...


5

I don't know if this is the real reason, but here's one possibility: because it's unnecessary. It's easy to say "JPEG = lossy = bad", but JPEG doesn't just destroy image quality willy-nilly. What the format gives you is a slider with "small, low-quality" at one end and "big, high-quality" at the other. You can easily reduce the byte size of a PNG image by ...


4

Lossy does not necessarily imply any major quality loss. It is mostly dropping out meaningless information of the image in favor a much smaller image file. The amount of quality drop can be adjusted, and in best case scenario this is done so that the quality drop is just invisible. In exchange for this quality drop you get quite big file size savings. And ...


2

It's dependent on the algorithm used for comparison, but if you have access to that algorithm, you can add small amounts of noise to pixels randomly until it misclassifies the image. If you had deeper knowledge of the algorithm used, and it was differentiable, you could use gradient descent (ascent) to modify the image to increase the error of what was ...


1

I would have done it with a script and NETPBM like so: 1) reducing Image1 (the tiles) to size X by X -> Image1R 2) recording the mapping from pixel values of Image1R to coordinates (no need to worry about duplicates as an exact RGB match of two or more averaged pixels is unlikely) 3) remapping Image2 using Image1R as the color map -> Image2M 4) reading ...


Only top voted, non community-wiki answers of a minimum length are eligible