Leveraging jpeg discrete cosine transform coefficients in neural networks

A non-transitory computer-readable storage medium storing instructions that when executed by a computer processor perform actions comprising:

• obtaining a compressed image file for analysis, wherein the compressed image file includes three channels of Huffman codes;
• decompressing the three channels of Huffman codes to obtain three blocks of quantized discrete cosine transform (DCT) values, wherein at least two of the blocks have differing dimensions;
• applying one or more transform functions to at least one of the three blocks of quantized DCT values, the one or more transform functions affecting the dimensions of the at least one of the three blocks such that the three blocks have the same dimensions;
• concatenating the resulting three blocks of values;
• and
• providing the concatenation of the values as input to a neural network to generate a classification of the compressed image file.

A computer system comprising:

• a computer processor;
• and
• a non-transitory computer-readable storage medium storing instructions that when executed by a computer processor perform actions comprising:
• obtaining a compressed image file for analysis, wherein the compressed image file includes three channels of Huffman codes;
• decompressing the three channels of Huffman codes to obtain three blocks of quantized discrete cosine transform (DCT) values, wherein at least two of the blocks have differing dimensions;
• applying one or more transform functions to at least one of the three blocks of quantized DCT values, the one or more transform functions affecting the dimensions of the at least one of the three blocks such that the three blocks have the same dimensions;
• concatenating the resulting three blocks of values;
• and
• providing the concatenation of the values as input to a neural network to generate a classification of the compressed image file.