Surveillance cameras are rarely in focus. This is caused by subjective settings during installation and on-going focus drift from building vibrations. Lack of proper use, and understanding, of focus correction has resulted in wildly varying results among enhancement experts. Unfortunately, this has increased the need for expensive testimony to defend inaccurate judgment-based results. The better method is to apply a mathematical corrective lens to restore the video back to its original quality.
For each focus or blur convolution, there is an equal inverse calculated through a Fourier analysis. Finite Impulse Response (FIR) Wiener filtering retains high frequency details at optimal efficiency. By applying this filter in a circular lens-style pattern, focus is significantly improved through a restorative process at optimal speed and accuracy. Since this process must be applied per pixel, it may take hours to correct a few minutes of video.
An alternate method is to intentionally blur an image using a wide Gaussian pattern, which is then subtracted from the original, with the majority of the weight given to the original. This type of Unsharp masking is called Local Contrast Enhancement (LCE) and requires minimal computing power. Due to the low quality nature of most surveillance video, Unsharp causes exaggerated contrast with thickened contour lines, and should only be used when FIR focus correction is impractical.
For both methods, optimal visual acuity occurs when changes in contrast are maximized. Automating this process ensures maximum benefits,
and our algorithm is well over 99% accurate to the optimal focus. Unfortunately, the
industry continues to rely upon judgment-based speculation by the software operator.