System and process for optimizing false alarm probability for histogram matching

A system for optimizing the overall false alarm probability when using histogram matching to recognize an object in an image of a scene comprising:

• at least one general purpose computing device;
• and
• a computer program comprising program modules executable by the at least one computing device, wherein the at least one computing device is directed by the program modules of the computer program to,
• compute the false alarm rate for a given set of parameters, wherein said computation comprises,
• generating a prototype histogram from a model image depicting an object being sought in an image,
• determining the set of all possible test histograms that can be formed from a background image of the scene given a prescribed number of bins and a maximum pixel count for the bins,
• determining the subset, of test histograms from said set of all possible test histograms, which will cause a false alarm, said false alarm occurring when a histogram formed from the background image matches the prototype histogram when that histogram does not actually represent the object being sought,
• determining the probability of occurrence of each individual test histogram that will cause a false alarm in said subset, and
• summing the individual false alarm probabilities of occurrence of all histograms in said subset of test histograms which will cause a false alarm to establish the overall false alarm probability;
• and
• adjust the parameters to bring the false alarm rate down.

A computer-readable memory for optimizing the overall false alarm probability when using histogram matching to recognize an object in an image of a scene comprising:

• a computer-readable storage medium;
• and
• a computer program comprising program modules stored in the storage medium, wherein the storage medium is so configured by the computer that it causes a computing device to,
• distort the pixels of a model image of the scene which contains the object being sought to approximate an image exhibiting substantially independent and identically distributed image,
• extract a portion of the distorted model image representing the object to create a distorted model image of the object and a distorted background image of the scene without the object,
• generate a prototype histogram from the distorted model image of the object given a prescribed number of bins and a maximum count for the bins,
• determine the set of all possible test histograms that can be formed from the distorted background image using the prescribed number of bins and the maximum count for the bins,
• determine the subset of test histograms from said set of all possible test histograms which will cause a false alarm, said false alarm occurring when a histogram formed from the distorted background image matches the prototype histogram when that histogram does not actually represent the object,
• determine the probability of occurrence of each individual test histogram that will cause a false alarm in said subset, and
• sum the individual false alarm probabilities of occurrence of all histograms in said subset of test histograms which will cause a false alarm to establish the overall false alarm probability.