Method for selecting JPEG quantization tables for low bandwidth applications

Claim 1

A machine for transmitting color images containing text characters, the machine comprising: a scanner configured to render a color image into color source image data representing the color image; a compression engine that compresses the source image data into compressed image data, the engine including: means for converting the source image data into transformed image data using a discrete cosine transform function, the transformed image data including one or more blocks of data (Y), each block including a plurality of data elements (Y(k,l)) corresponding to a respective transform basis vector, means for storing a multi-element quantization table, each quantization table element (Q(k,l)) corresponding to a respective transform basis vector, wherein each quantization table element (Q(k,l)) includes corresponding weighting factor (w(k,l)) that represents a number of bits allocated to encoding the quantization table element (Q(k,l)) and the value of w(k,l) is determined based on the importance of the quantization table element (Q(k,l)) in preserving elements of the text characters that enhance reading efficiency of the text, and wherein the value of w(k,l) is equal to a first predetermined value for quantization elements that correspond to transform basis vectors representing at least one of the `up-downness` and the `right-leftness` of the text characters, a second predetermined value for quantization elements that are adjacent to only one of the quantization elements having weighting factors equal to the first predetermined value, and a third predetermined value for quantization elements that are adjacent to two or more of the quantization elements having weighting factors equal to the first predetermined value, a quantizer that converts the transformed image data into quantized image data (Y Q ) in accordance with the quantization table elements, and an entropy encoder that transforms the quantized image data into the compressed image data; and means for transmitting the compressed image data.

Claim 18

A method of quantizing source image data using a discrete cosine transform function, the method comprising: receiving source image data representing a source image; selecting a threshold term that results in perceptually lossless quantization of the image; selecting a bit rate term so that the expected value of the compressed image size equals a predetermined size; selecting a supra-threshold term that results in perceptually lossy compression but where the supra-threshold term is selected to preserve critical textual elements; combining the threshold term, the bit rate term, and the supra-threshold term to form a corresponding quantization table element; dividing each source image datum by a respective quantization table element to form a quantized image datum.

Claim 19

A method of quantizing source image data using a discrete cosine transform function, the method comprising: receiving source image data representing a source image; selecting a threshold term that results in perceptually lossless quantization of the image; selecting a bit rate term so that the expected value of the compressed image size equals a predetermined size; selecting a supra-threshold term that results in perceptually lossy compression but where the supra-threshold term is selected to preserve critical textual elements; combining the threshold term, the bit rate term, and the supra-threshold term to form a corresponding quantization table element; dividing each source image datum by a respective quantization table element to form a quantized image datum.