DSP-based image real-time dehazing optimization for improved dark-channel prior algorithm

To solve the problem of non-real-time processing of image dehazing using traditional dark-channel prior algorithm, this work studies image real-time penetrating fog optimization technologies based on digital signal processor (DSP) devices. Using jointed optimization mechanism between algorithm and device, we can achieve real-time processing. During algorithm optimization, mean filter characterized low computation substitutes the guided filter which is the most complex in dark-channel algorithm for dehazing. In optimization of image processing task under the embedded device, we empirically construct two-step optimization strategy for raising speed of processing. Thereupon, the awful division calculation for DSP device is achieved approximately by multiplication after the reciprocal operation. We utilize the specified template which is considerably designed to realize mean filter. Thus, the division factor in the template can be calculated innovatively via shift instructions featured on DSP. The experimental results show that the optimization solution provided has realized real-time image dehazing processing for standard-definition and high-definition at frame rate of 25 fps over C6748 pure DSP device featured 456 MHz clock, at the same time the effect of penetrating fog is not remarkably degraded. The optimization methods or ideas can easily be transplanted to similar platform.