Generating custom quantization tables for jpeg compression based on image content

In a digital medium environment for compressing digital images, a method comprising:

• converting, by at least one processor, a plurality of blocks of a digital image to a frequency domain using a discrete cosine transform;
• creating, by the at least one processor, an averaged block representing the plurality of blocks in the frequency domain, the averaged block comprising a plurality of coefficients representing content in the digital image;
• determining, by the at least one processor, a plurality of compression error thresholds for the plurality of coefficients of the averaged block based on a location of each coefficient within the averaged block and a distribution of the plurality of coefficients within the averaged block;
• generating, by the at least one processor, a custom quantization table associated with the plurality of blocks by assigning a plurality of quantizers to the plurality of coefficients based on the plurality of compression error thresholds;
• and
• compressing, by the at least one processor, the digital image by applying the custom quantization table to the plurality of blocks.

In a digital medium environment for compressing digital images, a system comprising:

• at least one processor;
• and
• a non-transitory computer readable storage medium comprising instructions that, when executed by the at least one processor, cause the system to:
• identify, for each color space component from the plurality of color space components of a digital image, a plurality of blocks in a frequency domain;
• generate, for the plurality of color space components, a plurality of custom quantization tables by performing steps comprising:
• calculating a plurality of averaged coefficients across the plurality of blocks;
• determining, for each averaged coefficient from the plurality of averaged coefficients, a compression error threshold;
• and
• determining a plurality of quantizers that produce a plurality of compression errors that meet the compression error thresholds for the plurality of averaged coefficients;
• and
• compress the digital image using the plurality of custom quantization tables for the plurality of color space components.