: Why does JPEG make fonts and typography blurry? Which step in data compression / transformation makes it blurry? Is it the undersampling, the discrete cosinus transformation? quantization? or the

@Margaret771

In JPG the spectrum components (= the results of the cosine transform) are rounded to low bit resolution. That makes same amplitudes to repeat often and there also starts to occur larger repeating patterns in the sequence of spectrum components. For example weak spectrum components all become rounded to zero, when their non-zero bits are ignored from all numbers.

The existence of repeating patterns makes possible to use LZW or other compression methods that replace repeating patterns with shorter symbols. This is the way how JPG creates compression.

So, the rounding of the results of the cosine transform increase the possiblities to compression. Unfortunately it also blurs the image by adding unwanted patterns (=JPG artifacts) when the coarse spectrum is decoded to show the image. The artifacts spoil most effectively those parts of the image that need a rich and precise spectrum to be represented properly. Those parts are the sharp edges - such as in text.

Vote Up Vote Down