Comment by GrantS | Hacker News Reader

Comment by GrantS | Hacker News Reader

It's a strange post because nobody answers the question posed by the asker who is clearly a complete beginner, wondering how images get converted to grayscale and sepia -- I think it's a much more basic question than the answers imply.

If you already know how to code, the math at the top of this article answers the question. http://www.techrepublic.com/blog/howdoi/how-do-i-convert-ima...

If not: Digital images are a big grid of pixels. Each pixel has a red, green, and blue value, each of which is usually a number between 0 and 255 because 8 bits are used to represent each value.

A "black and white" image (really grayscale) is created by setting the red, green, and blue values at each pixel to be equal to the exact same value as each other, for example we could choose the average of the original red, green, and blue values at that pixel. This removes all "color".

Sepia is similar but you take three differently weighted averages of the original red, green, and blue values to produce the sepia-red, sepia-blue, and sepia-green values in the result. See above link for the specific weights.

In both cases, each pixel location is computed independently, whereas other operations, like blurring, need to look at other nearby pixels as well.

The first reply in that post tells that - "Essentially every image is nothing but a matrix of numbers, right? Numbers in the matrix can denote several things. For B/W image, these can be grey scale numbers from 1 to 255 (1 being the White and 255 being the Black). For colored images, each entry in the matrix can be a vector of (Red, Blue, Green) scales, since every color can be represented by these three basic colors.

Now, when you want to produce any effect, essentially you need to do some operation on this matrix. For example, if you want to remove the high contrast (sudden change of color from white to black), then you need to remove the high frequency components from your matrix. Taking a Fourier transform and removing the high frequency components and taking back the inverse transform can do the trick for you. I'm not sure about Sepia effect in particular, but all I want to convey here is doing some operation on Matrix does the job. Hope it helps."

Yes, though there is confusion in the second paragraph of that reply about what it means for the image to be a "matrix". It's being a matrix has no bearing on converting it to grayscale or sepia. It may as well be an unordered bag of pixels as far as those operations are concerned. I think most beginners would come away with the wrong impression that linear algebra or fourier transforms were important for these basic operations when one or two for loops is quite enough. When I teach this material, that's where I begin and it is very effective.

Also they use the term contrast incorrectly. They're describing a low pass filter (blur) in the example but refer to contrast -- that post would lead beginners very astray.

Actually, to get acceptable results even when converting from RGB to grayscale, you must weight the components unequally. See, for example, https://en.wikipedia.org/wiki/Grayscale#Converting_color_to_...

Reply on news.ycombinator.com