No abstract

Bad weather, for example fog, haze, smog, etc. Nice to advanced video surveillance will be enhanced, so the needs of pattern recognition AI and related research will increase. The problem is that the contrast of the photographed image decreases, the distant visibility deteriorates greatly, the components, size, and shape of fogs are different but the adverse effects are similar. This study is based on a large amount of analysis of predecessors on the Dark channel prior defogging algorithm based on the research results, especially the He team's research results, depth of field processing, to abandon the He team in order to enhance the sense of depth to set the range of depth of field approach. In this research, we proposed a method to remove fog from uneven defogging image with fog density change, using dark channel prior and local gradient key point extraction technology. Estimate the transmission map by using the image local information and shaping the coarse map using the detailed map. In this way, a clear image is obtained while maintaining the fog removing quality.

For the random noise of digital image processing is embedded in the old films, to study the improved Non Means algorithm in Local. Through the whole area of the surrounding pixels of gray distribution in contrast to the overall, according to the gray level distribution similarity obtained decision weights, to study and calculate the corresponding each pixel set color Information adaptive parameters, and then calculate the similarity, screened based on the template. In order to achieve the goal of eliminating the noise by modifying the value of the target pixel region, and using the weighted average value instead of the target pixel region value. Finally, the feasibility and the scientific nature of the noise reduction performance are simulated by referring to the image noise elimination precision and image quality parameters.