Blur difference estimation using multi-kernel convolution

An apparatus for automatically focusing an electronic image capturing device, comprising:

• a computer processor configured for processing images and controlling focus for an electronic image capturing device;
• and
• programming executable on said computer processor for adjusting focus by performing steps comprising:
• estimating actual subject distance in response to determining blur difference between subject images;
• modeling blur change as a point spread function from a first position to a second position, while determining said blur difference;
• approximating said point spread function as a series of convolutions,
• performing a first portion of said series of convolutions in response to utilizing a first kernel having a first variance;
• performing the remaining convolutions of said series of convolutions in response to utilizing at least a second kernel having at least a second variance which is smaller than said first variance;
• adjusting focus of the electronic image capturing device;
• and
• iteratively performing the above steps until a desired accuracy of focus is obtained.

A method for determining blur difference between at least a first and a second image, comprising:

• modeling blur change as a point spread function from a first position to a second position;
• approximating, on a computer processor, said point spread function as a series of convolutions;
• performing a first portion of said series of convolutions in response to using a first kernel having a first variance;
• and
• performing the remaining convolutions of said series of convolutions in response to using at least a second kernel having a second variance which is smaller than said first variance.