Image definition evaluation algorithm based on color relativity

Definition evaluation function of digital image plays an important role in digital camera auto-focus. Due to the existing definition evaluation functions are of some marked disadvantages, a spatial domain evaluation algorithm for digital color image was proposed. The colorful property of each pixel was judged and definition chromatic difference parameters was created by using chromatic difference between their tri-stimulus values. Besides,nonlinear function was used to improve the gradient coefficient of each pixel which made the evaluation function be more sensitive to some images in extreme cases. Experimental results show the superiority of our algorithm over the most of existing evaluation algorithms in dealing with natural images. And our method also is of a good ability of robustness as well as reducing calculation complexity and it can be easily achieved on hardware.