Method of compressing digital images acquired in CFA format

A method for compressing digital images acquired in CFA format and subjected to a processing sequence comprising at least a full-resolution expansion phase of the CFA image to generate a complete image, said method comprising the following operations

• decomposing the complete digital image into a plurality of planes (or channels),
• compressing each plane by means of a discrete transform into a two-dimensional spatial frequency domain, using for each plane a quantization matrix of which the elements are quantization coefficients that on the basis of their position are associated with particular spatial frequencies,
• characterized in that the quantization matrix is synthesized for at least one plane of the image by carrying out the following operations:
• selecting an arbitrary starting matrix Qst,
• assigning to each spatial frequency a statistical measure of the error introduced into said spatial frequency by the processing performed on the plane under consideration, and
• modifying the matrix Qst by multiplying each of its elements by an appropriate weighting coefficient that is a function of the statistical error measure assigned to the corresponding spatial frequency.

A method, comprising:

• generating a first image of a scene with an apparatus, the first image having a first set of pixels;
• generating a second image of the scene with a test apparatus, the second image having a second of pixels that are more accurate than the first set of pixels;
• transforming the first and second sets of pixels into respective first and second sets of transform coefficients;
• calculating the respective differences between the transform coefficients in the first set and the corresponding transform coefficients in the second set;
• calculating from the differences respective error-correction weights that correspond to the transform coefficients in the first set;
• and
• calibrating the apparatus to weight transform coefficients of a subsequent image with the respective error-correction weights.

A method, comprising:

• generating a first set of images of respective scenes with a camera, the first images each having a respective first set of pixels;
• generating a second set of images of the respective scenes with a test apparatus, the second images each having a respective second set of pixels that are more accurate than the corresponding first set of pixels;
• transforming each of the first and second sets of pixels into respective first and second sets of transform coefficients;
• calculating the average differences between the transform coefficients for the images in the first set and the corresponding transform coefficients for the images in the second set;
• calculating from each average difference a respective error value equal to the average difference divided by the sum of the average differences;
• calculating from each error value a respective weighting factor equal to a constant divided by the log2 of the error value;
• and
• calibrating the camera to multiply each quantization value used to compress transform coefficients of subsequent images with a respective one of the weighting factors.

A method, comprising:

• acquiring pixels of an image;
• transforming the pixels of the image into transform coefficients;
• and
• quantizing the transform coefficients with respective quantization values, at least one of the quantization values having been weighted according to a statistical error introduced to the corresponding at least one transform coefficient.